Natural Language Words Analysis for Affective Scene Generation from Written Text using Artificial Neural Network

Atul Deshkar¹, Avinash Dhole¹and Prafulla Vyas²

¹Raipur Institute of Technology, Madir Hasod, Raipur,

²Pt. Ravishankar Shukla University, Raipur (C.G.)

*Corresponding Author E-mail: atul_deshkar@yahoo.com; avi_dhole33@rediffmail.com; prafullavyas@gmail.com

ABSTRACT:

This paper presents an artificial neural network approach to word analysis to generate the 3 dimension scene or image from the textual description. We start with the recognition of characters and then form the words from these characters. The words used in natural language will have some special meaning and gives some information. Each word represents some inherited properties of some of the objects. The properties of each word will depend on the object being used in the sentence. Therefore the word itself gives lots of information about the objects. The neural network approach to lexical classifications is the first step to find the objects and its properties. The next step is neural network based approach to word classification is extracting words attribute and then relating it with the other words using artificial neural network. The multilayer feedforward neural network will be used. Here we will analyze the different parts of the speech with their inherit properties which the word have in the sentence.

A central issue in cognitive neuroscience today concerns how distributed neural networks in the brain that are used in language learning and processing can be involved in non-linguistic cognitive sequence learning. This issue is informed by a wealth of functional neurophysiology studies of sentence comprehension, along with a number of recent studies that examined the brain processes involved in learning non-linguistic sequences, or artificial grammar learning (AGL). The current research attempts to reconcile these data with several current neurophysiologically based models of sentence processing, through the specification of a neural network model whose architecture is constrained by the known cortico-striato- thalamo-cortical (CSTC) neuroanatomy of the human language system. The challenge is to develop simulation models that take into account constraints both from neuranatomical connectivity, and from functional imaging data, and that can actually learn and perform the same kind of language and artificial syntax tasks. Thus different distributed neural networks will be trained and integrated in such a way that it understands the language as being understand by the human being.

KEYWORDS: Part of speech, neural network, cognitive neuroscience, Computational Linguistic, Perception network.

INTRODUCTION:

Communication and perception of language consist of language representation, language acquisition, language understanding and language processing.

Thus different models have been proposed to learn language by the machine. The neural network based models will be proposed to understand the language as being understand by the human being. The distributed approach to neural network will be used in order to understand the language as neuro cognition language understanding. Thus character recognition of the language, word recognitions, semantic lexical information, artificial grammar recognition, object recognition, attributes of the words recognition based on the objects to be reference in the sentence, spatial position of the objects recognition based on the lexical properties of the words in the sentence[1-3]. The above networks will again be subdivided into different networks based on the different situations. Thus the simulating all these network in an integrated environment is a major and challenging task for machine work like human to recognize language.

The resulting neural network model could simulate human-like behavior in the learning of serial, temporal and abstract structure of language as revealed by experiments with human infants.[4]

In order to draw the scene from written natural language the following different modules are to be required. First, the recognition of character in that language. Secondly, words formed in that language by the combination of characters called lexicons. Third, the lexicon semantic analysis, object extractions, relationship of the objects with respect to the words seen in the sentence[5-6]. Relating spatial position of the objects with respect to each other. Extracting the proper object with described attributes in the sentence[7-8]. And then placing all the objects in the one figure where it shows the scene which is described in the written text. The neurocognition approach to the above modules are to be considered to develop human like understanding of language.

The flow diagram for this is shown below in the diagram.

Linguistic Analysis:

The combination of words are called sentence. The smallest logical unit in the sentence is called lexicon. The lexicon of the word gives some useful information to draw the scene. The English corpus is divided into different parts of speech[8-12]. The different parts-of-speech gives different information not only to the users but also to the computer system.

The words are classified into ten main categories and can be further be subdivided into 35 sub categories. The main categories are noun, pronoun, verb, adjective, adverb, preposition, article, conjunction, cardinal numbers and interjection[12-14]. These main categories may further be divided into 35 sub categories.

The human being recognized each character, recognizes words and then lexical meaning to each of these words. Thus one by one the human understands and learns words and understand the meaning associated with the words. For example Boy is a singular common noun and it recognizes a single boy having different features or properties like shape, gesture, color, height etc[16]. The human understands each and every object by seeing shape of the object it looks like. Thus the internal representation of the shape of an object is formed inside the brain and thus whenever the same word appears it extracts that impression of the brain and thus recognizes the object. Therefore the neural network approach is used for language understanding.

Figure 1. Steps to generate Scene from a written text.

The construction of a novel phrase detector capable of automatically finding phrases in an unstructured English text corpus. After training, the detector can be exposed to an arbitrary string of English text and break it up into its component phrases. This is the first step in constructing a higher-level set of networks capable of abstracting the phrases into unitary tokens, and identifying the relationship between phrases.

Subjectivity is a pragmatic, sentence-level feature that has important implications for text processing applications such as information extraction and information retrival. The study of adjectives and semantically oriented adjectives are strong predictors of subjectivity.

The user describes the environment that they wish to create using adjectives. An entire scene is then procedurally generated, based on the mapping of these adjectives to the parameter space of the procedural models used.

Adjective space and parameter space

Suppose that the user can choose adjectives from a set A to describe their VE. We define adjective space, A, to be a subset of jAj-dimensional real values, A = [􀀀1; 1] jAj (where jAj denotes the number of adjective in the set A). Each dimension in A thus relates to a unique adjective in A, and the value x in any dimension of A is a real value in the range [􀀀1; 1] which is the scalar value associated with the relevant adjective. Hence an element of A describes a specific set of scalar values associated with the adjectives in A, and A represents the set of all possible descriptions that a user could make.[2]

Spatial Relations [20]

A system based on neural networks that can analyse spatial relations in a visual scene and connect them to appropriate linguistic descriptions. The system learns spatial concepts like “right of” and “above” by viewing a visual scene containing a number of objects and simultaneously receiving a text string describing the scene. The system thus learns to correlate linguistic expressions for spatial relation with different kinds of saccades. After being trained, the system can correctly describe previously viewed scenes. In our simulations the linguistic units have been nouns and prepositional phrases. Our approach is based on extractive picturable keyword summarization. That is, it builds on standard keyword-based text summarization

Semantic Representations [21]

There are two types of semantic representations: One for displaying objects with their attributes and the other for displaying spatial relationships between two objects in the scene. The first type of representations is as: Class:OBJ(name, count, s_x, s_y, s_z, c_r, c_g, c_b, material), where, name is the object’s name, count is the quantity of that object, and s_x, s_y, s_zare scaling parameters all in x, y and z directions. c_r, c_g, c_bdetermine the red, green and blue color components. Material determines what that object is made of. Class in the representation of the objects is the most specific ontological concept describing that object in our ontology.

The second type of representations is as:

• ABOVE(object₁; object₂)

• BELOW(object₁; object₂)

• ON(object₁; object₂)

• UNDER(object₁; object₂)

• INSIDE(object₁; object₂)

• AROUND(object₁; object₂)

• RIGHT(object₁; object₂)

• LEFT(object₁; object₂)

• FRONT(object₁;object₂)

• BEHIND(object₁; object₂)

Where, object₁and object₂are the objects that are related to each other spatially.

Action State

The verb phrases gives action and shape of the object at that instance of time. It defines the tense or happening of the events.

Knowledge Representation

As each word contains attributes and gives inherit information. Each word should maintain the knowledge and attributes it have and described it in scalar quantity. This will help in extracting the appropriate object from the database based on the relationships of the words.

Neural Networks Approach

An artificial neural network is a mathematical model or computational model that is inspired by the structure and/or functional aspects of the biological neural system. It consists of an interconnected group of artificial neurons and process information using a connectionist approach to computation. In most cases an ANN is an adaptive system that changes its structure based on external or internal information that flows through the network during the learning phase. Modern neural networks are non linear statistical data modeling tools. They are usually used to model complex relationships between inputs and outputs or to find patterns in data. In the present work NN approach has been adopted in parsing the text and object recognition using classification. Artificial neurons are the constitutive units in an artificial neural network. It receives one or more inputs and sums them to produce an output. Usually the sums of each node are weighted, and the sum is passed through a non linear function known as an activation function or transfer function. The transfer function is usually have a sigmoidal, piecewise linear functions or step functions. They are also often monotonically increasing, continuous, differentiable and bounded.

Advantages of Neural Network:

· A neural network can perform tasks that a linear program cannot.

· When an element of the neural network fails, it can continue without any problem by their parallel nature.

· A NN learns and does not need to be reprogrammed.

· It can be implemented in any application.

· It can be implemented without any problem.

Disadvantages of Neural Network:

· The neural network needs training to operate.

· The architecture of a neural network is different from the architecture of microprocessors therefore needs to be emulated.

· Requires high processing time for large neural network.

Perceptron Network Architecture

The perceptron network consists of a single layer of S perceptron neurons connected to R inputs through a set of weights w^i,j, as shown below in two forms. As before, the network indices i and j indicate that wi,j is the strength of the connection from the j^th input to the i^th neuron.

The perceptron learning rule described shortly is capable of training only a single layer. Thus only one-layer networks are considered here. This restriction places limitations on the computation a perceptron can perform.

The different perceptron network can prepared for different modules as defined in above linguistic analysis. The different networks are trained and are interrelated with each other. This simulation is the crucial task that has to be performed in order to simulate like human language understanding.

The different levels of analysis of the lexical semantics are performed to find the exact meaning of the sentence.

CONCLUSION:

In order to simulate all these features we need different modules hence different neural networks will be trained to perform the specific tasks being defined in each module. They are distributed networks each connected with some type of relations based on the words in the sentence. This will help in simulating the human like language understanding mechanism. The lexical classification for natural language using perceptron neural network had already developed and gives a very good results. The scene generation is the next step of lexical classification. Hence multilayered feedforward perceptron neural network is in progress and expected to give acceptable results as seen in the preliminary stage of the project. This paper is just the review of different language features used in analyzing the language words features in order to develop the machine which generates scene from the text written in natural language using neural cognition network.

REFERENCE:

1. Lars Kopp. A neural network for spatial relations: connecting visual scenes to linguistic descriptions, Lund University Cognitive Science, Kungshuset, Lundagård, S–222 22 LUND, Sweden.

2. Frank L. Borchardt. Neural Network Computing and Natural Language Processing Duke University.

3. KIMURA Kazuhiro SUZUOKAT akashi AMANO Sin-ya. Association-based Natural Language Processing with Neural Networks. Information Systems Laboratory, Research and Development Center, TOSHIBA Corp., 1 Komukai-T6siba-ty6, Saiwai-ku, Kawasaki 210 Japan.

4. Peter Ford Dominey a,*, Toshio Inui b,c, Michel Hoen. Neural network processing of natural language: II. Towards a unified model of corticostriatal function in learning sentence comprehension and non-linguistic sequencing, a INSERM U846 Stem Cell and Brain Research Institute, 18 ave. Doyen Lepine, 69675 Bron Cedex, France, b Department of Intelligence Science and Technology, Graduate School of Informatics, Kyoto University, Kyoto 606-8501, Japan c ERATO Asada Project, Japan Science and Technology Agency, Japan

5. Deepyaman Maiti1, Ayan Acharya2, Amit Konar3A. Novel Parser Design Algorithm Based on Artificial Ants, Department of Electronics and Telecommunication Engineering, Jadavpur University, Kolkata: 700032, India.

6. Sergio Roa and F. Nino. Classification of Natural Language Sentences using Neural Networks. National University of Colombia, Dept. of Comp. Sci. Ciudad University, Bogota, Columbia.

7. A. Rajkumar and Dr. A. Chitra. Paraphrase Recognition using Neural Network Classification, International Journal of Computer Applications: 2010:Volume 1-No.29.

8. Hehnut Schmid. Part-Of-Speech Tagging With Neural Networks. Institute for Computational Linguistics, Azenbergstr.12, 70174 Stuttgart, Germany.

9. Richard C. Murphy. Phrase Detection and the Associative Memory Neural Network. Computer Science and Engineering Department, University of Notre Dame, rcm􀀀 cse.nd.edu.

10. Word Net the Lexical Database.

11. Michael Collins. Head-Driven Statistical Models for Natural Language Parsing. MIT Computer science and Artificial Intelligence Laboratory.

12. H. Svensson and Ola Aerberg. Development and Integration of Linguistic Components for an Automatic text to Scene Conversion System. Lund Institute of Technology, August 2002.

13. K. Church. A Stochastic parts program and noun phrase parser for unrestricted text. Bell Laboratories, In proceedings of 2nd Conference on Applied Natural Language Processing, Morristown, NJ,. Association for Computational Linguistics. 1988:pages 136-143.

14. C. Hultquist and J. Gain. Affective Scene Generation. Department of Computer Science, University of Cape Town, South Africa.

15. X. Zhu, A.B. Goldberg. A Text to Picture Synthesis System for Augmenting Communication. Department of Computer Science, University of Wiscosin,Madison, WI, USA.

16. Bob Coyne and Richard Sproat . Words Eye: An automatic Text to Scene conversion system. AT&T Labs .paper presented at SigGraph Annual meeting 2001, Los Angeles, CA.

17. John Funge, Xiaoyuan tu and D. Terzopoulos. Cognitive Modeling: Knowledge,Reasoning and planning for intelligent Characters. Intel Corporation Univeersity of Toronto. In SIGGRAPH 99, Conference proceedings, ACM SIGGRAPH, Addison Wesley, 1999: pages 29-38.

18. Dan Tappan. Towards Automated scene generation from textual descriptions. (New Mexico State University).

19. M. Arens and A. Ottlik and H. Nagel. Natural Language Texts for a Cognitive Vision System.

20. S. R. Clay and j. Wilhelms. Put: Language-Based Interactive Manipulation of Objects. IEEE Computer Graphics and Applications, march 1996: pages 31-39.

21. B. Hawkins. The semantics of English Spatial Prepositions. PhD thesis, University of California, San Diego, CA, 1984.

22. Norbert E. Fuchs, Rolf Schwitter. Attempto Controlled Natural Language for Requirements Specifications. University of Zurich, CH- 8057, Zurich, Switzerland