RICH SHORT TEXT CONVERSATION USING SEMANTIC KEY CONTROLLED SEQUENCE GENERATION

if you want the project pls call @8125424511

RICH SHORT TEXT CONVERSATION USING SEMANTIC KEY CONTROLLED SEQUENCE GENERATION

ABSTRACT:

Through series-line-up improvements Generation approaches to configuration, short text dialog (STC) becomes attractive. Classical sequence-line-line approaches Short text conversations often suffer from poor people Common answer without distinction. It's hard Control the title or semantics of the selected record Many created candidates. On this sheet, a novel exterior A continuous learning approach from memory driven line has been proposed Face these problems. External memory is an audit Built to represent understandable topics or semantics. When Generation, given a controlled memory stimulus Input array, and an answer created using memory later on Trigger and serial-line sequence model. Experiments The proposed approach shows that they can create the most rich The difference between traditional sequencing-line sequence training Attention. Meanwhile, it achieves excellent quality in man Evaluation. This is done by manually manipulating it Memory trigger, headings can guide directly Answer or semantic explanation.

ARCHITECTURE:

EXISTING SYSTEM:

With the widespread usage of social media, such as Twitter and microblogs, in recent years, more and more open domain conversation data becomes feasible, which makes data-driven approach for conversation possible. Short Text Conversation (STC)  is a simplified conversation task: one round conversation formed by two short text sequences. It is widely used in conversation robot for chit-chat. The former one, usually given by human being, is referred to as a post, while the latter, given by computer, is referred to as a comment. The research on STC contributes to the development of open domain conversation system. There are two major frameworks for short text conversation: retrieval based methods and generation based methods. Retrieval based methods search the STC training corpus to find an existing comment which is most relevant to the post. Generation methods  usually train a text generation model on the STC corpus and generate a comment using the model given a post. Compared to retrieval based methods, generation based methods can produce new comments that are not in the training set. This important feature makes generation methods very attractive.

DISADVANTAGAE:

1.  Again and again user asks the same question for the admin.

2.  Admin maintains the data base is difficultly. Training data set is     update the very lengthy process.

3.  Waste of time for in this process.

PROPOSED SYSTEM:

The encoder part encodes the variable length sequence into a fixed length vector. Then, the decoder part generates a variable length sequence from this vector word by word. Although this method successfully links variable length input and output into a single model, it suffers from vanishing gradient problem when the input is too long. In addition, a fixed length vector can not encode sufficient information when the input is long. Attention mechanisms  have been proposed to tackle this problem. When generating the next word, the decoder can access all hidden vectors of the encoder. Then, the decoder network decides which segment of the input is more relevant to the current situation by computing a soft alignment. The alignment is a byproduct of the sequence-to-sequence training.The vector is then used, as an auxiliary feature, together with the post sentence embedding to be input to the decoder during training and generation. By enumerating different semantic keywords extracted from the post, it is possible to generate comments with rich diversity. Moreover, it is even possible to manually manipulated memory trigger process to introduce new semantics which does not exist in the post.In this work, we combine the advantages of  andand propose a new sequence-to-sequence learning approach for STC. A tensor, in the form of a list of matrices, is constructed to represent the semantics of the comment sentences, referred to as external semantic memory. Each matrix represents all possible comment sentences corresponding to a specific semantic key. Each row vector of the matrix forms a sentence embedding basis and all row vectors span the whole comment semantic space of the specific semantic key. During generation, a semantic key is extracted from the input sequence and used to construct a comment sentence embedding from the external memory. The final comment is then generated using the embedding from external memory as well as the post sequence embedding with a sequence-to-sequence model. By manipulating the semantic keys, it is possible to interpretably guide the topics or the semantics of the comment.

ADVANTAGE:

1. Response the same question is avoided for the admin.

2. Easy for text classification is simply process.

3.Time consume is save for the admin.

MODULES:

1.   User Activity

User login the our system or mobile phone or any text devices. Our clarify questions and response the admin the every time in same questions but not answer the same answer. The admin reply for the user.

2.   Training Dataset

User send the text conversation store the database and admin maintaining the training data set or back of words .In back of words  using for data mining concepts. This process is Mainly used for admin reply save consume the time process.

3.   Text Classification

The data mining process or text classifications using for respited answer in reply in the admin. The admin very excite the reply for the one or many times in the same questions or answer. The classification is the one main process in training data set is comparing the user text conversation.

4.   GRAPHICAL NOTATIONS

The collected data are representing as graphical form which is help to identify the best way of analyzing the performance of proposed system. The graphs are different like pie chart, bar chart and column chart. The better way to understand the data in which it helps to find the best method among available.

ALGORITHM:

Data Mining Algorithm

Data Mining is a technique used in various domains to give mean-ing to the available data. In classification tree modeling the data is classified to make predictions about new data. Using old data to pre-dict new data has the danger of being too fitted on the old data.In this paper we compare different data mining methods and techniques for classifying students based on their Moodle usage data and the final marks obtained in their respective courses.

Support Vector Machine

“Support Vector Machine” (SVM) is a supervised machine learning algorithm which can be used for both classification or regression challenges. However, it is mostly used in classification problems. In this algorithm, we plot each data item as a point in n-dimensional space (where n is number of features you have) with the value of each feature being the value of a particular coordinate. Then, we perform classification by finding the hyper-plane that differentiate the two classes very well (look at the below snapshot). The SVM algorithm is implemented in practice using a kernel. The learning of the hyperplane in linear SVM is done by transforming the problem using some linear algebra, which is out of the scope of this introduction to SVM. A powerful insight is that the linear SVM can be rephrased using the inner product of any two given observations, rather than the observations themselves. The inner product between two vectors is the sum of the multiplication of each pair of input values. For example, the inner product of the vectors [2, 3] and [5, 6] is 2*5 + 3*6 or 28. The equation for making a prediction for a new input using the dot product between the input (x) and each support vector (xi) is calculated as follows:

f(x) = B0 + sum(ai * (x,xi))

This is an equation that involves calculating the inner products of a new input vector (x) with all support vectors in training data. The coefficients B0 and ai (for each input) must be estimated from the training data by the learning algorithm.

REQUIREMENT ANALYSIS

          The project involved analyzing the design of few applications so as to make the application more users friendly. To do so, it was really important to keep the navigations from one screen to the other well ordered and at the same time reducing the amount of typing the user needs to do. In order to make the application more accessible, the browser version had to be chosen so that it is compatible with most of the Browsers.

REQUIREMENT SPECIFICATION

Functional Requirements

§  Graphical User interface with the User.

Software Requirements

For developing the application the following are the Software Requirements:

1.    Python

2.    Django

3.    MySql

4.    MySqlclient

5.    WampServer 2.4

Operating Systems supported

1.    Windows 7

2.    Windows XP

3.    Windows 8

Technologies and Languages used to Develop

1.    Python

Debugger and Emulator

§  Any Browser (Particularly Chrome)

Hardware Requirements

For developing the application the following are the Hardware Requirements:

§  Processor: Pentium IV or higher

§  RAM: 256 MB

§  Space on Hard Disk: minimum 512MB

CONCLUSION:

This paper proposes a new generation approach for short text conversation task. By incorporating external semantic memory in encoder-decoder framework, the approach greatly alleviates the problem of general replies without substantiality and generates more diverse and concrete responses. Both objective evaluation and human evaluation demonstrate the advantages of this new approach. The separation of external memory construction and neural network training also makes it possible to utilize non-parallel corpora. Furthermore, the semantics of generated responses can be controlled by manipulating the semantic key mapper, which implies a new way to generate rich responses. Due to data scarcitys, when data driven mapping functions like embedding or random mapping are used, the systems may generate incomprehensible comments, which is a problem to be addressed in the future.

Author : ss infotech

Share this

Related Posts