Title of Invention

A METHOD FOR CHATTING

Abstract

The present invention provides a method for chatting including: acquiring language information inputted, analyzing main sentence elements of the language information and determining syntax, and determining the topic attribute of the language information according to a preset topic dictionary, the main sentence elements and the syntax; determining whether there is the sentence pattern corresponding to the main sentence elements and the syntax, selecting an answer corresponding to the topic attribute from default answers and feeding back the selected answer if no, and determining a pattern, selecting an answer from the answers corresponding to the pattern and feeding back the selected answer if yes. The present invention also provides a system for chatting. The system includes: an acquiring unit, a natural language processing unit, a computational language knowledge storing unit, a topic understanding unit, a topic dictionary storing unit, a reasoning unit, a reasoning knowledge storing unit and a sending unit.

Full Text

Method and System for chatting Field of the Technology The present invention relates to methods of interactive conversation on the Internet, and particularly, to a method and system for chatting. Background of the Invention The Internet has experienced stupendously fast development and becomes a part of daily lives of modern people. One of the basic functions of the Internet is to provide mutual communications of people over the Internet. For example, both the initial e-mail and the current popular instant messaging tools possess large user groups. Another major function of the Internet is to provide information. However, the amount of the information on the Internet is tremendous and it is very difficult to find the desired information even if there are strong searching tools. In addition, modern people is subjected to faster and faster living rhythm as well as greater and greater pressure, and at the same time, modern people have less and less communions with others. Thus, some chatting systems based on an instant messaging tool or short message platform have come into existence to meet the demand of changes in social development. Such a system is used for performing man-machine conversation through natural language words, the core of which is a natural language processing technology and which relates to multiple subjects such as computational linguistics, information science and artificial intelligence. On the one hand, the system analyses and processes a question raised by a user, and on the other hand, the system generates a correct answer according to a repository. Thus, the system may have a chat with the user and perform multiple functions such as querying. For example, the system may perform the following two conversations with a user. Conservation 1 User: How about the weather in Shenzhen tomorrow? Robot: Clear and 26-32 centigrade in Shenzhen tomorrow. User: Great. Robot: A good weather for outing. Conservation 2 User: What programs will be played in the spring festival evening? Robot: Some comic dialogues and skits will be played and Liu Dehua will present his new song. User: Great. Robot: The springing festival evening must be wonderful this year. The same saying, "Great", is inputted by the user in the two conversations. However, the user said "Great" in conversation 1 after knowing the weather, while said "Great" in conversation 2 after knowing the programs to be played in the spring festival evening. A user may say "Great" in many other situations. But it is easy to find, according to the context language environments, that the meaning of "Great", which the user says, is different, A chatting system is implemented on the basis of the technology in which sentence pattern matching is performed according to the natural language information inputted by a user and then the corresponding answer is returned. However, it is a prevalent phenomenon that the topics a user talks with the system have the same sentence pattern, but the topic attributes may not be identical. The pattern in the present invention means the attributes, such as syntax and refined elements extracted by processing the natural language information inputted by a user. Therefore, some labels associated with the topic attributes of users are designed artificially in the answers to be returned when the repositories are constructed in some chatting systems. These attributes are identical with the attributes of sentences in the chatting system. After receiving and understanding the natural language information inputted by a user, the chatting system matches the understood natural language information with a pattern in the system, searches out the answers matching the pattern, selects one of the answers, feeds the answer back to the user, and returns the label of topic attribute of the answer to the user at the same time. However, the chatting system can not know a new topic attribute when the system does not search out the pattern matching the information inputted by a user. Thus, it cannot be determined whether the topic attribute of the user has been changed, and one answer must be selected and fed back to the user according to the original topic attribute. Accordingly, the situation in the following conversation easily occurs. User: How about the weather in Shenzhen tomorrow? Robot: Clear and 26-32 centigrade in Shenzhen tomorrow. User: Shenzhen has succeeded in applying for holding the University Games. Robot: A good weather for outing. "Shenzhen has succeeded in applying for holding the University Games" inputted by the user in the above conversation is not matched with a pattern in the system and thus the last topic attribute, the topic of weather, is continued. However, in the above method, since the topic attribute is configured when the user inputs a sentence and the chatting system returns an answer for the last time, in the case that the user inputs a new question sentence for this time and the new sentence does not match any sentence pattern in the repository, the topic attribute of the new question sentence cannot be distinguished, the topic attribute of the last sentence is continued, and a default answer is selected according to the original topic attribute. However, the topic attribute for this time may be transferred in fact, which can not provide a user with a better experience. Summary of the Invention The objective of the present invention is to provide a method and system for chatting continuously. The present invention provides the following technical solutions: acquiring natural language information inputted, analyzing main sentence elements of the natural language information and determining syntax, and determining a topic attribute of the natural language information according to a preset topic dictionary, the main sentence elements and the syntax; determining whether there is a sentence pattern corresponding to the main sentence elements and the syntax; selecting an answer corresponding to the topic attribute from default answers and feeding back the selected answer, if no; determining the sentence pattern, selecting an answer from the answers corresponding to the sentence pattern and feeding back the selected answer, if yes. An embodiment of the present invention also provides a system for chatting corresponding to the method. The system includes: an acquiring unit, a natural language processing unit, a computational language knowledge storing unit, a topic understanding unit, a topic dictionary storing unit, a reasoning unit, a reasoning knowledge storing unit and a sending unit; the acquiring unit is configured to acquire natural language information inputted by a user; the natural language processing unit is configured to analyze the natural language information according to computational language knowledge, and determine main sentence elements and syntax of the natural language information; the computational language knowledge storing unit is configured to store knowledge necessary for understanding the natural language information; the topic understanding unit is configured to distinguish the topic attribute of the natural language information inputted by the user according to the main sentence elements, the syntax and a topic dictionary; the topic dictionary storing unit is configured to store the topic dictionary including association relation between key words and topics in natural language information; the reasoning unit is configured to determine, according to reasoning knowledge, whether there is a sentence pattern corresponding to the main sentence elements and the syntax, select an answer corresponding to the topic from the default answers if no, and otherwise, determine a sentence pattern and selecting an answer from the answers corresponding to the sentence pattern; the reasoning knowledge storing unit is configured to store various sentence patterns matching the natural language information inputted by the user, and answers with topic attribute and default answers with topic attribute corresponding to each of the sentence patterns; the sending unit is configured to feed the answer selected by the reasoning unit back to the user. The present invention also provides a method for chatting including: acquiring inputted natural language information, analyzing main sentence elements of the natural language information and determining syntax of the natural language information, distinguishing the topic attribute of the natural language information according to the main sentence elements and the syntax and determining the topic attribute of the natural language information acquired previously for the at least one time as the topic attribute for this time if the natural language information has no theme; determining whether there are multiple topic attributes; selecting a sentence pattern matching the natural language information from preset sentence patterns if there are no multiple topic attributes; selecting an answer from the answers corresponding to the sentence pattern and outputting the answer if a sentence pattern is selected; selecting an answer from the default answers and outputting the answer if no sentence pattern is selected; selecting a sentence pattern matching the natural language information from the preset sentence patterns for each of the multiple topic attributes, selecting a sentence pattern from the selected sentence patterns randomly, and selecting an answer from the answers corresponding to the sentence pattern and outputting the answer, if there are multiple topic attributes; selecting an answer from the default answers and outputting the answer if there is no sentence pattern matching the natural language information. The present invention also provides a system for chatting corresponding to the method. The system includes: an acquiring unit, a natural language processing unit, a topic understanding unit, a reasoning unit, an sending unit, a computational language knowledge storing unit, a topic dictionary storing unit, and a reasoning knowledge storing unit; the acquiring unit is configured to acquire natural language information inputted by a user; the natural language processing unit is configured to analyze the natural language information according to computational language knowledge and determine main sentence elements of the natural language information and syntax of the natural language information; the computational language knowledge storing unit is configured to store knowledge necessary for understanding the natural language information; the topic unit is configured to distinguish the topic attribute of the natural language information according to the main sentence elements, the syntax and a topic dictionary, and determine the topic attribute of the natural language information acquired previously for the at least one time as the topic attribute for this time if the natural language information has no theme, and determining whether there are multiple topic attributes; the topic dictionary storing unit is configured to store a topic dictionary including association relation between key words and topics in natural language information; the reasoning pattern and answer unit is configured to select a sentence pattern matching the natural language information from preset sentence patterns in the case that there are not multiple topic attributes, select an answer from the answers corresponding to the sentence pattern if the sentence pattern is selected; select an answer from the default answers if no sentence pattern is selected; and select a sentence pattern matching the natural language information from the preset sentence patterns for each of the multiple topic attributes, select a sentence pattern from the selected sentence patterns randomly, and select an answer from the answers corresponding to the sentence pattern if there are multiple topic attributes; select an answer from the default answers if there is no sentence pattern matching the natural language information; the reasoning knowledge storing unit is configured to store various sentence patterns matching the natural language information inputted by the user, and answers with topic attribute and default answers with topic attribute corresponding to each of the sentence patterns; the sending unit is configured to feed an answer selected by the reasoning unit back to the user. It can be seen in the above technical solutions that the probability that a topic is unable to be distinguished due to no matched pattern is reduced since the process of distinguishing the topic attribute is performed before a pattern is matched. Moreover, the topic attribute of the last sentence is further taken for the topic attribute of a new sentence when it is determined that the topic attribute of the new sentence is having no theme. Thus, the conversation can be performed continuously when the new sentence has no theme. Brief Description of the Drawings Fig. 1 shows a schematic diagram illustrating the multi-level topic hierarchy in accordance with an embodiment of the present invention. Fig. 2 shows a flow chart of the chatting method with multi-level topics in accordance with an embodiment of the present invention. Fig. 3 shows a diagram illustrating the chatting system with multi-level topics in accordance with an embodiment of the present invention. Fig. 4 shows a flow chart of the chatting method with uni-level topics in accordance with an embodiment of the present invention. Fig. 5 shows a diagram illustrating the chatting system with uni-level topics in accordance with an embodiment of the present invention. Embodiments of the Invention Embodiments of the present invention provide a method and system for chatting on the basis of the distinguishing of a topic attribute. The embodiments are hereinafter described in detail. As shown in Fig. 1, a class hierarchy for multi-level topic mechanism, i.e., a multi-level topic hierarchy based on the granularity of topics, is to be established first before a chat is performed with a user. Such a class hierarchy for multi-level topic mechanism is established according to a human knowledge structure. For example, the topic of movie and television stars is a topic of large granularity which includes multiple topics of the same little granularity, such as the topic of Liu Dehua and the topic of Zhang Xueyou. A computational language repository for processing a natural language is necessary to be established first since much training is required in processing the natural language and some statistical relations need to be established. The repository includes but is not limited to various statistical data, such as a dictionary, word frequencies, grammar rules and semantic rules, with which information such as main sentence elements and syntax may be acquired. Moreover, a statistical relation between key words and topics and a mapping relation between key words and topics of the minimum granularity, both of which are multiple-to-multiple relations in general, also need to be established. A key word has different scores for different topics and a repository with correlation probabilities between key words and topics is established to form a topic dictionary. A reasoning repository needs to be established on the basis of the establishment of the above basic contents. The reasoning repository stores a plurality of contents such as sentence patterns corresponding to various natural language possibly inputted by a user as well as answers and default answers corresponding to each of the sentence patterns. The answers are also categorized according to the granularity of multi-level topics. Referring to Fig. 2, a method for chatting with a user is hereinafter described in detail. Step 201: Acquire natural language information inputted by a user, analyze main sentence elements and determine syntax. The analyzing is performed by using the contents stored in a computational ' language repository in this step. The contents are an important branch of research in the field of artificial intelligence which enables a computer to understand and apply a human natural language so as to achieve natural language based effective communications between a human being and a computer. The natural language in the present invention means various natural languages used by human beings such as Chinese and English. Step 202: Distinguish a topic attribute. Since natural language information inputted by a user is protean, the topic attribute can be distinguished with information such as the main sentence elements and the syntax sometimes, while the topic attribute cannot be distinguished sometimes. For example, the topic attributes of sentences such as "Great" cannot be distinguished. Thus, the distinguished topic attribute includes that the topic attribute is having a theme and the topic attribute is having no theme. If a user inputs a sentence such as "Great", the theme of the conversations of the previous multiple times with the user is determined as the theme of this topic. Step 203: Determine whether there is a conflict between the topics, perform Step 204 if yes, and otherwise, perform Step 205. The conflict between the topics in this step means that multiple themes of topic of the same level, i.e., the same granularity, may be included in the natural language information inputted by the user and thus the system cannot distinguish which theme of topic is talked by the user to the end, since a class hierarchy for multi-level topic mechanism is established in accordance with the embodiment of the present invention. Step 204: Enlarge the granularity and determine a topic attribute. Specifically, enlarge the granularity for the multiple themes of topic of the same granularity, search the topics in the case of the enlarged granularity, i.e., a level which is upper than the level of the granularity, and determine whether a theme of topic of larger granularity, which contains all the multiple themes of topic of the same* granularity, can be searched out. If such a theme of topic is searched out, the theme of topic of larger granularity is determined as the theme of topic of this conversation. If such a theme of topic is not searched out, this conversation is determined as having no theme. Step 205; Determine whether there is a corresponding sentence pattern according to the analyzed main sentence elements, the determined syntax and a reasoning repository preset in the system, perform Step 206 if yes, and otherwise, perform Step 217. The criterion for determining whether there is the corresponding sentence pattern is high grade of matching. Only the pattern of which the grade of matching reaches a certain threshold may be determined as a pattern corresponding to the main sentence elements and the determined syntax. Step 206: Determine whether only one sentence pattern is matched, perform Step 208 if yes, and otherwise, perform Step 207. Step 207: Since the topic attribute is further determined as having a theme or having no theme in Step 204, a sentence pattern is arbitrarily selected from multiple sentence patterns in the case that the topic attribute is determined as having no theme upon enlarging the granularity level by level The sentence pattern matching the topic is selected from multiple sentence patterns in the case that the topic attribute is determined as having a theme upon enlarging the granularity level by level or the theme is determined when there is no conflict between the topics. Step 208: Determine whether the sentence pattern has a theme of topic, perform Step 209 if yes, and otherwise, perform Step 210. Step 209: Determine whether there is an answer to topics of minimum granularity matching the topic in the answers corresponding to the sentence pattern having a theme of topic, perform Step 214 if yes, and otherwise, perform Step 211. The answers stored in the reasoning repository is also categorized according to multi-level granularity of topic, and the answers to the topics of minimum granularity are first searched in order to return an answer closely matching the theme of topic to the user. Step 210: Select an answer arbitrarily from the answers corresponding to the sentence pattern and feed the answer back to the user. Step 211: Determine whether the granularity of topic is of the uppermost level, i.e. the largest granularity, according to the preset hierarchy for multi-level granularity of topic, perform Step 212 if no, and perform Step 210 if yes. Step 212: Enlarge the granularity. The operation is this step is similar to that in Step 204. That is, enlarge the granularity for the multiple themes of topic of the same granularity, search the topics in the case of the enlarged granularity, i.e., a level which is upper than the level of the granularity, and determine whether a theme of topic of larger granularity, which contains all the multiple themes of topic of the same granularity, can be searched out. Step 213: Determine whether there is an answer matching the topic in the case of the granularity of the upper level, perform Step 219 if no, and perform Step 214 if yes. Step 214: Determine whether there are multiple answers matching the topic, perform Step 215 if yes, and otherwise, perform Step 216. Step 215: Select an answer arbitrarily from the multiple answers matching the topic. Step 216: Feed the selected answer back to the user. Step 217: Determine whether the topic attribute determined in Step 204 is having a theme when there is no matched sentence pattern, perform Step 218 if yes, and otherwise, perform Step 222. Step 218: Determine whether there is an answer matching the topic of minimum granularity in the default answers, perform Step 214 if yes, and otherwise, perform Step 219, The default answers stored in the reasoning repository are also categorized according to multi-level granularity of topic, and the answers to the topics of minimum granularity are first searched in order to return an answer closely matching the theme of topic to the user. Step 219: Determine whether the granularity of topic is of the uppermost level, i.e. the largest granularity, according to the preset hierarchy for multi-level granularity of topic, perform Step 220 if no, and perform Step 222 if yes. Step 220: Enlarge the granularity. The operation is this step is similar to that in Step 204. That is, enlarge the granularity for the multiple themes of topic of the same granularity, search the topics in the case of the enlarged granularity, i.e., a level which is upper than the level of the granularity, and determine whether a theme of topic of larger granularity, which contains all the multiple themes of topic of the same granularity, can be searched out. Step 221; Determine whether there is an answer matching the topic in the case of the granularity of the upper level, perform Step 219 if no, and perform Step 214 if yes. Step 222: Select an answer arbitrarily from all the default answers and feed the answer back to the user. On the base of the embodiment of the present invention, the process of determining a theme of topic may be further fined, and the corresponding processing is performed when a theme of topic is transferred, in order to further improve the continuity of conversation and thus improve satisfactory of users. The method is substantially the same as the above embodiments except that corresponding changes are to be made to Steps 202 and 204. Therefore, only the parts of the method in Steps 202 and 204 are hereinafter described in detail, and specifically, the corresponding contents in the above embodiment are to be replaced with the following ones. Other steps will not be described. Step 202: Distinguish the topic attribute and determine whether the topic is transferred. Since natural language information inputted by a user is protean, the topic attribute can be distinguished with information such as the main sentence elements and the syntax sometimes, and the topic attribute cannot be distinguished sometimes. For example, the topic attributes of sentences such as "Great" cannot be distinguished. Thus, the distinguished topic attribute includes that the topic attribute is having a topic and the topic attribute is having no topic. If a user inputs a sentence such as "Great", the theme of the conversations with the user for the previous multiple times is determined as the theme of this topic. When the topic attribute is distinguished, it is determined whether the topic is transferred, i.e., it is determined, by comparing the topic for this time with the topics for the previous multiple times, whether the topic for this time is transferred, and then the result of the determination is stored, Step 204; Enlarge the granularity and determine the topic attribute. Specifically, enlarge the granularity for the multiple themes of topic of the same granularity, search the topics in the case of the enlarged granularity, i.e., a level which is upper than the level of the granularity, and determine whether a theme of topic of larger granularity, which contains all the multiple themes of topic of the same granularity, can be searched out. If such a theme of topic is searched out, the theme of topic of larger granularity is determined as the theme of topic of this conversation. If such a theme of topic is not searched out, the corresponding processing is performed according to the result of the determination of topic transfer in Step 202, Specifically, the theme of the previous topic is determined as the theme of topic of this conversation if the topic is not transferred, while the theme of this conversation is determined as having no theme if the topic is transferred. Since the storage capacity of the system is finite, an active topic transfer may be performed in accordance with an embodiment of the -present invention in order to enable the system to have a more fluent conversation with a user. For example, in accordance with the above embodiment, the active topic transfer may be implemented with the following method. Enlarge the granularity for the topic attribute in the case that there are sentence patterns corresponding to the main sentence elements and the syntax but the number of the sentence patterns matching the topic is less than a preset threshold. Determine, on the basis of the enlarged granularity of topic attribute, whether the number of the sentence patterns of other topic attributes with the same granularity of topic attribute as that of the topic attribute for this time exceeds the preset threshold, and select the topic attribute having the most preset contents from the topic attributes. The topic attribute is transferred to a new topic and the answer to the transferred topic attribute is fed back, if the topic attribute is selected; an answer corresponding to the topic attribute is selected from the default answers and is fed back, if the topic attribute is not selected. An embodiment of the present invention also provides a system for chatting based on the distinguishing of a topic attribute, referring to Fig. 3. The system includes: an acquiring unit, a natural language processing unit, a computational language knowledge storing unit, a topic understanding unit, a topic dictionary storing unit, a reasoning unit, a reasoning knowledge storing unit and a sending unit. The acquiring unit is configured to acquire natural language information inputted by a user. The natural language processing unit is configured to analyze the natural language information according to computational language knowledge and determine the main sentence elements and the syntax of the natural language information. The computational language knowledge storing unit is configured to store knowledge used for understanding the natural language information. The topic understanding unit is configured to distinguish the topic attribute of the natural language information inputted by a user according to the main sentence elements, the syntax and a topic dictionary. The topic dictionary storing unit is configured to store a topic dictionary including association relation between key words and topics in natural language information. The reasoning unit is configured to determine, according to reasoning knowledge, whether there is a sentence pattern corresponding to the main sentence elements and the syntax, select an answer corresponding to the topic from the default answers if no, and otherwise, determine a sentence pattern and select an answer from the answers corresponding to the sentence pattern. The reasoning knowledge storing unit is configured to store various sentence patterns matching natural language information inputted by a user, and answers with topic attribute corresponding to each of the sentence patterns and default answers with topic attribute. Each sentence pattern corresponds to answers with topic attribute and the topic attribute of a pattern is a set of the topic attributes of all the answers. The reasoning knowledge storing unit also includes default answers to which no pattern corresponds. The default answers also have topic attributes. The sending unit is configured to feed an answer selected by the reasoning unit back to the user. The topic understanding unit specifically includes: a topic distinguishing unit, a topic granularity determining unit, a first topic selecting unit, a second topic selecting unit and a topic transfer determining unit. The topic distinguishing unit is configured to distinguish whether the topic attribute of the acquired natural language information is having no theme. The first topic selecting unit is configured to set the topic attribute for last time as the topic attribute for this time according to the result of having no theme distinguished by the topic distinguishing unit. The second topic selection unit is configured to enlarge, according to the result that there are multiple topics of the same granularity determined by the topic granularity determining unit, the granularity level by level till a topic containing all the multiple topics of the same granularity is selected; determine that the topic attribute for this time is having no theme if the topic containing all the multiple topics of the same granularity is not selected, determine the theme of the topic for the last time as the theme of topic for this time in the case that it is determined that the topic is not transferred, there are multiple topics of a same granularity, and the topic attribute for this time is determined as having no theme after the granularity is enlarged level by level, and determine that the topic for this time is having no theme in the case that it is determined that the topic is transferred, there are multiple topics of the same granularity, and the topic attribute for this time is determined as having no theme after the granularity is enlarged level by level The topic granularity determining unit is configured to determine whether the determined topic attribute of minimum granularity for this time contains the multiple topics of the same granularity. The topic transfer determining unit is configured to determine, by comparing the theme of the topic for this time with the themes of topics for the previous multiple times, whether the theme of the topic for this time is transferred, in the case that the topic distinguishing unit distinguishes that the topic attribute for this time has a theme. The topic dictionary storing unit is configured to store a topic dictionary including association relation between key words and topics in natural language information. The reasoning unit specifically includes: a pattern determining unit, a default answer selecting unit, a minimum granularity determining unit, a default answer topic selecting unit, a first no-theme answer selecting unit, a first answer selecting unit, a second no-theme answer selecting unit, a second answer selecting unit, a threshold determining unit, a topic transferring unit and a third answer selecting unit. The pattern determining unit is configured to determine, according to reasoning knowledge, whether there is a sentence pattern corresponding to the main sentence elements and the syntax. The default answer selecting unit is configured to select an answer from default answers when there is no matched sentence pattern and the second topic selecting unit determines that the topic attribute is having no theme. The minimum granularity determining unit is configured to determine whether there is an answer matching the topic of the minimum granularity in default answers, when there is no matched sentence pattern and the second topic selecting unit determines that a topic attribute is having a theme. The default answer topic selecting unit is configured to select an answer from the answers matching the topic when it is determined that there is an answer matching the topic of the minimum granularity in default answers, and select an answer from the default answers when it is determined that there is not an answer matching the topic of the minimum granularity in the default answers. The first no-theme answer selecting unit is configured to select an answer from the answers corresponding to the sentence pattern when the pattern determining unit determines that there is only one sentence pattern corresponding to the main sentence elements and syntax and the topic attribute is having no theme. The first answer selecting unit is configured to select an answer matching the topic from the answers corresponding to one sentence pattern when the pattern determining unit determines that there is only the sentence pattern corresponding to the main sentence elements and syntax and the topic attribute is having a theme. The second no-theme answer selecting unit is configured to select an answer from the answers corresponding to a sentence pattern matched most closely when the pattern determining unit determines that there are multiple sentence patterns corresponding to the main sentence elements and the syntax and the topic attribute is having no theme. The second answer selecting unit is also configured to select an answer matching the topic attribute from the answers corresponding to the sentence pattern matching the topic attribute when the pattern determining unit determines that there are multiple sentence patterns corresponding to the main sentence elements and the syntax and the topic attribute is having a theme. The reasoning knowledge storing unit is further configured to store the answers with the transferred topic attribute corresponding to each sentence pattern. The threshold determining unit is configured to determine whether there is a sentence pattern corresponding to the main sentence elements and the syntax but the number of sentence patterns matching the topic is less than a preset threshold, and enlarge the granularity of the topic attribute, and determine, on the basis of the enlarged granularity of the topic attribute, whether the number of the sentence patterns of the other topic attributes with the same granularity of topic attribute as that of the topic attribute for this time exceeds the preset threshold. The topic transferring unit is configured to transfer the topic attribute to a new topic when the number of sentence patterns of other topic attributes with the same granularity as that of the topic attribute for this time exceeds a preset threshold. The third answer selecting unit is configured to select an answer from the answers with the transferred topic attribute corresponding to each sentence pattern when the topic attribute is transferred. The above embodiment is performed in the case of multi-level topic mechanics while another embodiment is performed in the case of uni-level topic mechanics. That is, all the topics are of the same granularity, referring to Fig. 4. The specific procedure of the embodiment includes the following steps. Step 401; Acquire natural language information inputted by a user, and analyze main sentence elements and determine syntax. The analyzing is performed by using the contents stored in a computational language repository in this step. The contents are an important branch of research in the field of artificial intelligence which enables a computer to understand and apply a human natural language so as to achieve natural language based effective communications between a human being and a computer. The natural language in the present invention means various natural languages used by human beings such as Chinese and English. Step 402: Distinguish a topic attribute. Since natural language information inputted by a user is protean, the topic attribute can be distinguished with information such as the main sentence elements and the syntax sometimes while the topic attribute cannot be distinguished sometimes. For example, the topic attributes of sentences such as "Great" cannot be distinguished. Thus, the distinguished topic attribute includes that the topic attribute is having a topic and the topic attribute is having no topic. If a user inputs a sentence such as "Great", the theme of the conversations of the previous multiple times with the user is determined as the theme of this topic. Step 403: Determine whether there is a conflict between the topics, perform Step 410 if yes, and otherwise, perform Step 404. The conflict between the topics in this step means that multiple themes of topic of the same level, i.e., the same granularity, may be included in the natural language information inputted by the user and thus the system cannot distinguish which theme of topic is talked by the user to the end, since a uni-level topic hierarchy is established in the embodiment of the present invention. Step 404: Determine whether there are corresponding sentence patterns according to the analyzed main sentence elements, the determined syntax and a reasoning repository preset in the system, perform Step 405 if yes, and otherwise, perform Step 413. The criterion for determining whether there are the corresponding sentence patterns is high grade of matching. Only the sentence pattern of which the grade of matching reaches a certain threshold may be determined as a sentence pattern corresponding to the main sentence elements and the determined syntax. Step 405: Determine whether there is only one matched sentence pattern, perform Step 407 if yes, and otherwise, perform Step 406. Step 406: Select a sentence pattern matching the topic attribute from the sentence patterns, perform Step 407 if the sentence pattern is selected, and select a sentence pattern arbitrarily from the sentence patterns and perform Step 407 if the sentence pattern is not selected. Step 407: Determine whether there is an answer matching the topic in the answers corresponding to the sentence pattern, perform Step 409 if there is not the answer, and perform Step 408 if there is the answer. Step 408: Select an answer arbitrarily from the answers matching the topic and feed the answer back to the user. Step 409: Select an answer arbitrarily from the answers corresponding to the sentence pattern and feed the answer back to the user. Step 410: Perform sentence pattern matching for all the topics where there is a conflict. Step 411: Determine whether there is a matched sentence pattern, perform Step 412 if yes, and otherwise, perform Step 413. Step 412: Select a question sentence pattern arbitrarily from all the sentence patterns where there is a conflict and perform Step 407. Step 413: Determine whether there is an answer matching the topic in default answers, perform Step 414 if yes; otherwise, perform Step 415. Step 414: Select an answer arbitrarily from the answers matching the topic corresponding to the sentence pattern and feed the answer back to the user. Step 415: Select an answer arbitrarily from all the default answers corresponding to the sentence pattern and feed the answer back to the user. The present invention also provides another system for chatting, referring to Fig, 5. The system includes: an acquiring unit, a natural language processing unit, a topic unit, a reasoning pattern and answer unit, a sending unit, a computational language knowledge storing unit, a topic dictionary storing unit, and a reasoning knowledge storing unit. The acquiring unit is configured to acquire natural language information inputted by a user. The natural language processing unit is configured to analyze the natural language information according to computational language knowledge and determine main sentence elements and syntax of the natural language information. The computational language knowledge storing unit is configured to store knowledge used for understanding the natural language information. The topic unit is configured to distinguish the topic attribute of the natural language information according to the main sentence elements, the syntax and a topic dictionary, and determine the topic attribute for the last time as the topic attribute for this time if the language information has no theme, and determine whether there are multiple topic attributes. The topic dictionary storing unit is configured to store a topic dictionary including association relation between key words and topics in natural language information. The reasoning pattern and answer unit is configured to select a sentence pattern matching the natural language information from preset sentence patterns in the case that there are not multiple topic attributes, select an answer from the answers corresponding to a sentence pattern if the sentence pattern is selected; select an answer from the default answers if no sentence pattern is selected; and select a sentence pattern matching the natural language information from the preset sentence patterns for each of the multiple topic attributes, select a sentence pattern from the selected sentence patterns randomly, and select an answer from the answers corresponding to the sentence pattern, in the case that there are multiple topic attributes; select an answer from the default answers if there is no sentence pattern matching the language information. The reasoning knowledge storing unit is configured to store various sentence patterns matching the natural language information inputted by the user, answers with topic attribute corresponding to each of the sentence patterns and default answer with topic attribute. The sending unit is configured to feed an answer selected by the reasoning unit back to the user. It can be seen in the above technical solutions that the probability that a topic is unable to be distinguished due to no matched pattern is reduced since the process of distinguishing the topic is performed before a pattern is matched* Moreover, the topic attribute of the former sentence is further taken for the topic attribute of the sentence for this time when it is determined that the topic attribute of a sentence is having no theme. Thus, the conversation can be performed continuously when the topic attribute has no theme. When it is determined that the topic attribute is having a theme, the theme for this time is further compared with the topic attributes for the previous multiple times to determine whether the topic is transferred, and the corresponding processing is performed if the topic is transferred and it is determined with the granularity based method that the topic attribute is having no theme. Thus, the continuity of conversation is further improved and satisfactory of users is improved. Since the storage capacity of the system is finite, in order to have a more fluent conversation with a user, an active topic transfer may be performed in accordance with an embodiment of the present invention, which enables the system to have a conversation with a user more intelligently. In accordance with embodiments of the present invention, a better experience may be provided to users by returning a corresponding default answer according to the current topic attribute when a pattern cannot be matched in the chatting system, and a user may be correspondingly led to a neighboring topic attribute according to the distribution of knowledge classes in the repositories, which ensures a more fluent man-machine conversation* Moreover, it is determined whether the topic attribute is to be continued or transferred according to the language information inputted by a user so that the attribute of the topic that the user talks currently may be distinguished better, and a better experience may be provided to the user by selecting the default answer corresponding to the current topic attribute preferably and returning the selected answer when the new sentence does not match any sentence pattern. Therefore, a method for determining whether the topic attribute is to be transferred and a process of understanding the information and recognizing the topic attribute after a user inputs natural language information are implemented in accordance with the embodiments of the present invention to solve the problem of the evaluation of a topic attribute after pattern matching in the prior art. The method and system for chatting in accordance with the embodiments of the present invention are described in detail above. The principles and implementation methods of the present invention are explained with specific examples in the present invention. The description of the above embodiments is only for use in helping understanding the methods of the present invention and the core idea thereof. Meanwhile, for those skilled in the art, changes may be made to both the specific implementation method and application scope based on the idea of the present invention. To sum up, this description should not be regarded as limiting the present invention. Claims 1. A method for chatting, comprising: acquiring natural language information inputted, analyzing main sentence elements of the natural language information and determining syntax, and determining a topic attribute of the natural language information according to a preset topic dictionary, the main sentence elements and the syntax; determining whether there is a sentence pattern corresponding to the main sentence elements and the syntax; selecting an answer corresponding to the topic attribute from default answers and feeding back the selected answer, if no; determining the sentence pattern, selecting an answer from the answers corresponding to the sentence pattern and feeding back the selected answer, if yes. 2. The method of Claim 1, wherein the determining the topic attribute of the natural language information comprises: distinguishing whether the natural language information has no theme; determining the topic attribute of the natural language information acquired previously for at least one time as the topic attribute of the natural language information acquired for this time, if the natural language information has no theme; determining, according to a preset topic granularity structure, whether there are multiple topic attributes of the same granularity in the case of the minimum 1 granularity of the topic attribute of the natural language information acquired for this time; determining whether there is a sentence pattern corresponding to the main sentence elements and the syntax in preset sentence patterns, if there are not the multiple topic attributes; • enlarging the minimum granularity according to the preset topic granularity structure and selecting a topic attribute containing all the multiple topic attributes of the same granularity, before determining whether there is a sentence pattern corresponding to the main sentence elements and the syntax, if there are the multiple topic attributes; determining that the natural language information tor this time nas no tneme, n the topic attribute containing all the multiple topic attributes of the same granularity is not selected. 3. The method of Claim 2, further comprising: determine whether the topic attribute is transferred by comparing the theme of the topic attribute with a theme of topic attribute of the natural language information acquired previously for at least one time, when it is distinguished that the topic attribute for this time has a theme; determining the topic attribute of the natural language information acquired previously for at least one time as the topic attribute of the natural language information acquired for this time in the case that there are multiple topic attributes of the same granularity, it is determined, after the granularity is enlarged, that the topic attribute is having no theme, and the topic attribute is not transferred; determining the topic attribute for this time as having no theme in the case that there are multiple topic attributes of the same granularity, it is determined, after the granularity is enlarged, that the topic attribute is having no theme, and the topic attribute is transferred. 4, The method of any of Claims 1 to 3, wherein the selecting an answer corresponding to the topic attribute from default answers and feeding back the selected answer comprises; determining whether the topic attribute for this time is having no theme, selecting an answer from the default answers and feeding the answer back if the topic attribute for this time is having no theme; determining whether there is an answer matching the topic attribute of minimum granularity in the default answers if the topic attribute for this time is having a theme; feeding the answer back if there is the answer; enlarging the granularity of topic attribute, selecting an answer matching the topic attribute of the larger granularity and feeding the answer back, if there is not the answer; selecting an answer from the default answers and outputting the answer if the answer matching the topic attribute is not selected till the largest granularity is reached. 5. The method of Claim 1, wherein the answer is selected from the answers corresponding to the sentence pattern and fed back, if there is only one sentence pattern corresponding to the main sentence elements and the syntax and the topic attribute is having no theme; the answer matching the topic attribute is selected from the answers corresponding to the sentence pattern and is fed back, if the topic attribute is having a theme. 6. The method of Claim 1, wherein the sentence pattern matching the topic attribute closely is selected from the multiple sentence patterns and the answer is selected according to the sentence pattern matched closely and is fed back, if there are multiple sentence patterns corresponding to the main sentence elements and the syntax and the topic attribute is having no theme; the sentence pattern matching the topic attribute is selected and the answer matching the topic attribute is selected from the answers corresponding to the sentence pattern and is fed back, if there are multiple sentence patterns corresponding to the main sentence elements and the syntax, and the topic attribute is having a theme. 7. The method of Claim 5 or 6, wherein the selecting an answer matching the topic attribute from the answers corresponding to the sentence pattern if the topic attribute is having a theme comprises: determining whether there is an answer matching the topic attribute of the minimum granularity in the corresponding answers, feeding the answer back if there is the answer, and enlarging the granularity of topic attribute and selecting an answer matching the topic attribute of the larger granularity and feed the answer back, if there is not the answer; selecting an answer from the answers corresponding to the sentence pattern and outputting the answer, if the answer matching the topic attribute is not selected till the largest granularity is reached* 8. The method of Claim 5 or 6, wherein the granularity is enlarged for the topic attribute, it is determined, on the basis of the enlarged granularity of the topic attribute, whether the number of the sentence patterns of other topic attributes with the same granularity as that of the topic attribute for this time exceeds the preset threshold, and the topic attribute having the most preset contents is selected from the topic attributes if the number of the sentence patterns of which exceeds the preset threshold, in the case that there is the sentence pattern corresponding to the main sentence elements and the syntax but the number of the sentence patterns matching the topic is less than the preset threshold; and the topic attribute is transferred to a new topic and the answer to the transferred topic attribute is fed back, if the topic attribute is selected; an answer corresponding to the topic attribute is selected from the default answers and is fed back, if the topic attribute is not selected. 9. A system for chatting, comprising: an acquiring unit, a natural language processing unit, a computational language knowledge storing unit, a topic understanding unit, a topic dictionary storing unit, a reasoning unit, a reasoning knowledge storing unit and a sending unit; the acquiring unit is configured to acquire natural language information inputted by a user; the natural language processing unit is configured to analyze the natural language information according to computational language knowledge, and determine main sentence elements and syntax of the natural language information; the computational language knowledge storing unit is configured to store knowledge necessary for understanding the natural language information; the topic understanding unit is configured to distinguish the topic attribute of the natural language information inputted by the user according to the main sentence elements, the syntax and a topic dictionary; the topic dictionary storing unit is configured to store the topic dictionary comprising association relation between key words and topics in natural language information; the reasoning unit is configured to determine, according to reasoning knowledge, whether there is a sentence pattern corresponding to the main sentence elements and the syntax, select an answer corresponding to the topic from the default answers if no, and otherwise, determine a sentence pattern and selecting an answer from the answers corresponding to the sentence pattern; the reasoning knowledge storing unit is configured to store various sentence patterns matching the natural language information inputted by the user, and answers with topic attribute and default answers with topic attribute corresponding to each of the sentence patterns; the sending unit is configured to feed the answer selected by the reasoning unit back to the user. 10. The system of Claim 9, wherein the topic understanding unit comprises: a topic distinguishing unit, a topic granularity determining unit, a first topic selecting unit, and a second topic selecting unit; the topic distinguishing unit is configured to distinguish whether the topic attribute of the acquired natural language information is having no theme; the first topic selecting unit is configured to set the topic attribute for the last time as the topic attribute for this time according to the result of having no theme distinguished by the topic distinguishing unit; the second topic selection unit is configured to enlarge, according to the result determined by the topic granularity determining unit that there are multiple topics of the same granularity, the granularity of a topic containing all the topics of the same granularity is selected, and determine the topic attribute for this time as having no theme if the topic containing all the topics of the same granularity is not selected; the topic granularity determining unit is configured to determine, according to a preset topic granularity structure, whether there are multiple topics of the same granularity in the case of the minimum granularity of the determined topic attribute for this time. 11. The system of Claim 10, wherein the topic understanding unit further comprises: a topic transfer determining unit; the topic transfer determining unit is configured to determine, by comparing a theme of topic for this time with themes of topics for the previous multiple times, whether the theme of topic for this time is transferred, when the topic distinguishing unit distinguishes that the topic attribute has a theme; the second topic selecting unit is further configured to determine the theme of topic for the last time as the theme of topic for this time when it is determined that the topic is not transferred, there are multiple topics of the same granularity, and the topic attribute for this time is determined as having no theme after the granularity is enlarged, and determine the topic for this time as having no theme when it is determined that the topic is not transferred, there are multiple topics of the same granularity, and the topic attribute for this time is determined as having no theme after the granularity is enlarged. 12. The system of any of Claims 9 to 11, wherein the reasoning unit comprises; a pattern determining unit, a default answer selecting unit, a minimum granularity determining unit, and a default answer topic selecting unit; the pattern determining unit is configured to determine, according to reasoning knowledge, whether there is a sentence pattern corresponding to the main sentence elements and the syntax; the default answer selecting unit is configured to select an answer from default answers in the case that there is no matched sentence pattern and the second topic selecting unit determines that the topic attribute is having no theme; the minimum granularity determining unit is configured to determine whether there is an answer matching the topic of the minimum granularity of the topic attribute in default answers, in the case that there is no matched sentence pattern and the second topic selecting unit determines that a topic attribute is having a theme; the default answer topic selecting unit is configured to select an answer from the answers matching the topic when it is determined that there is an answer matching the topic of the minimum granularity of the topic attribute in default answers, and select an answer from the default answers when it is determined that there is not an answer matching the topic of the minimum granularity of the topic attribute in the default answers. 13. The system of Claim 9, wherein the reasoning unit further comprises: a first no-theme answer selecting unit and a first answer selecting unit; the first no-theme answer selecting unit is configured to select an answer from the answers corresponding to a sentence pattern when the pattern determining unit determines that there is only the sentence pattern corresponding to the main sentence elements and the syntax and the topic attribute is having no theme; the first answer selecting unit is configured to select an answer matching the topic from the answers corresponding to the sentence pattern when the pattern determining unit determines that there is only the sentence pattern corresponding to the main sentence elements and the syntax and the topic attribute is having a theme. 14. The system of Claim 9, wherein the reasoning unit further comprises: a second no-topic answer selecting unit, and a second answer selecting unit; the second no-theme answer selecting unit is configured to select an answer from the answers corresponding to the sentence pattern matching the topic attribute closely when the pattern determining unit determines that there are multiple sentence patterns corresponding to the main sentence elements and the syntax and the topic attribute is having no theme; the second answer selecting unit is further configured to select an answer matching the topic attribute from the answers corresponding to the sentence pattern matching the topic attribute when the pattern determining unit determines that there are multiple sentence patterns corresponding to the main sentence elements and the syntax and the topic attribute is having a theme. 15. The system of Claim 13 or 14, wherein the reasoning unit further comprises: a threshold determining unit, a topic transferring unit and a third answer selecting unit; the reasoning knowledge storing unit is further configured to store the answers with the transferred topic attribute corresponding to each of the sentence patterns; the threshold determining unit is configured to determine whether there is a sentence pattern corresponding to the main sentence elements and the syntax but the number of the sentence patterns matching the topic is less than a preset threshold; enlarge the granularity of the topic attribute, and determine, on the basis of the enlarged granularity of the topic attribute, whether the number of the sentence patterns of other topic attributes of the same granularity as that of the topic attribute for this time exceeds the preset threshold; the topic transferring unit is configured to transfer the topic attribute to a new topic when the number of sentence patterns of other topic attributes of the same granularity as that of the topic attribute for this time exceeds the preset threshold; the third answer selecting unit is configured to select an answer from the answers with the transferred topic attribute corresponding to each of the sentence patterns when the topic attribute is transferred. 16. A method for chatting, comprising: acquiring inputted natural language information, analyzing main sentence elements of the natural language information and determining syntax of the natural language information, distinguishing the topic attribute of the natural language information according to the main sentence elements and the syntax, and determining the topic attribute of the natural language information acquired previously for the at least one time as the topic attribute for this time if the natural language information has no theme; determining whether there are multiple topic attributes; selecting a sentence pattern matching the natural language information from preset sentence patterns if there are no multiple topic attributes; selecting an answer from the answers corresponding to the sentence pattern and outputting the answer if a sentence pattern is selected; selecting an answer from the default answers and outputting the answer if no sentence pattern is selected; selecting a sentence pattern matching the natural language information from the preset sentence patterns for each of the multiple topic attributes, selecting a sentence pattern from the selected sentence patterns randomly, and selecting an answer from the answers corresponding to the sentence pattern and outputting the answer, if there are multiple topic attributes; selecting an answer from the default answers and outputting the answer if there is no sentence pattern matching the natural language information. 17. The method of Claim 16, wherein the selecting a sentence pattern matching the natural language information from the preset sentence patterns comprises: determining whether there is a sentence pattern matching the natural language information in present sentence patterns; determining whether there is only one matched preset sentence pattern if there is a sentence pattern matching the natural language information in the present sentence patterns; selecting an answer from the answers corresponding to the sentence pattern and outputting the answer if there is only one matched preset sentence pattern; selecting a sentence pattern matching the topic attribute from the multiple matched preset sentence patterns if there are multiple matched preset sentence patterns; selecting an answer from the default answers and outputting the answer if there is not a sentence pattern matching the natural language information in the present sentence patterns. 18. The method of Claim 16 or 17, wherein the selecting an answer from the default answers and outputting the answer comprises: determining whether there is an answer matching the topic attribute in the default answers, selecting an answer randomly if no, and selecting an answer from the answers matching the topic attribute and outputting the answer if yes. 19. The method of Claim 16 or 17, wherein the selecting an answer from the answers corresponding to the sentence pattern comprises: determining whether there is an answer matching the topic attribute in the corresponding answers, selecting an answer randomly from the answers matching the topic attribute and outputting the answer if yes, and selecting an answer from all the corresponding answers and outputting the answer if no. 20, A system for chatting, comprising: an acquiring unit, an natural language processing unit, a topic unit, a reasoning pattern and answer unit, an sending unit, a computational language knowledge storing unit, a topic dictionary storing unit, and a reasoning knowledge storing unit; the acquiring unit is configured to acquire natural language information inputted by a user; the natural language processing unit is configured to analyze the natural language information according to computational language knowledge and determine main sentence elements of the natural language information and syntax of the natural language information; the computational language knowledge storing unit is configured to store knowledge necessary for understanding the natural language information; the topic unit is configured to distinguish the topic attribute of the natural language information according to the main sentence elements, the syntax and a topic dictionary, and determine the topic attribute of the natural language information acquired previously for the at least one time as the topic attribute for this time if the natural language information has no theme, and determining whether there are multiple topic attributes; the topic dictionary storing unit is configured to store a topic dictionary comprising association relation between key words and topics in natural language information; the pattern reasoning and answer unit is configured to select a sentence pattern matching the natural language information from preset sentence patterns in the case that there are not multiple topic attributes, select an answer from the answers corresponding to the sentence pattern if the sentence pattern is selected; select an answer from the default answers if no sentence pattern is selected; and select a sentence pattern matching the natural language information from the preset sentence patterns for each of the multiple topic attributes, select a sentence pattern from the selected sentence patterns randomly, and select an answer from the answers corresponding to the sentence pattern if there are multiple topic attributes; select an answer from the default answers if there is no sentence pattern matching the natural language information; the reasoning knowledge storing unit is configured to store various sentence patterns matching the natural language information inputted by the user, and answers with topic attribute and default answers with topic attribute corresponding to each of the sentence patterns; the sending unit is configured to feed an answer selected by the reasoning unit back to the user.

Documents:

721-CHE-2008 FORM-1 27-02-2014.pdf

721-CHE-2008 AMENDED PAGES OF SPECIFICATION 27-02-2014.pdf

721-CHE-2008 CORRESPONDENCE OTHERS 08-11-2012.pdf

721-CHE-2008 ENGLISH TRANSLATION 27-02-2014.pdf

721-CHE-2008 EXAMINATION REPORT REPLY RECEIVED 27-02-2014..pdf

721-CHE-2008 POWER OF ATTORNEY 27-02-2014.pdf

721-CHE-2008 AMENDED CLAIMS 20-12-2013.pdf

721-CHE-2008 CORRESPONDENCE OTHERS 08-11-2012.pdf

721-CHE-2008 EXAMINATION REPORT REPLY RECIVED 20-12-2013.pdf

721-CHE-2008 FORM-3 20-12-2013.pdf

721-CHE-2008 OTHER PATENT DOCUMENT 20-12-2013.pdf

721-che-2008-abstract.pdf

721-che-2008-claims.pdf

721-che-2008-correspondnece-others.pdf

721-che-2008-description(complete).pdf

721-che-2008-drawings.pdf

721-che-2008-form 1.pdf

721-che-2008-form 18.pdf

721-che-2008-form 3.pdf

721-che-2008-form 5.pdf