Information Extraction Using Neural Networks

Huge amounts of unstructured text are generated every day in the present digital era by books, research articles, websites, and social media. One of the biggest challenges is collecting valuable information from this expanding information pool. The process of automatically locating and classifying pertinent information from unstructured data, including people, locations, dates, and events, is known as information extraction (IE). IE systems facilitate the analysis, retrieval, and application of information by transforming disorganized text into structured knowledge. Neural networks, with their capacity to recognize patterns and comprehend context, have transformed the process, making Internet Explorer more precise and versatile across various domains. Previously, rule-based models frequently failed to handle the intricacies of natural language.
The human brain served as the model for neural networks, which are computer systems. They are made up of layers of linked nodes, or "neurons," that use weighted connections to process data.

The network can identify patterns and relationships because the weights are changed during training to reduce errors. Neural networks are used in information extraction to tokenize sentences, categorize words, and identify relationships, allowing machines to comprehend meaning that goes beyond simple grammar. The system can recognize the individual, place, and date, for instance, if the sentence is "Dr. S.R. Ranganathan was born in Tamil Nadu on 12 August 1892." A network stores this structured representation, which can subsequently be queried to deliver precise responses to user inquiries.

Neural networks are used in a multi-step information extraction process. The input document is first tokenized, giving each word a distinct identity. Words are then categorized by the system into categories like person, place, date, or event. Semantic roles, such as agent, location, or time, are then described by deep cases. A knowledge network is then created using the relationships that have been established between entities, where nodes stand in for entities and edges for connections. The system searches the network, evaluates keywords, and outputs the appropriate unit—in this case, Tamil Nadu—when a user queries it with a question like, "Where was Ranganathan born?"

Neural network-based IE applications are numerous. They extract metadata from digital libraries, including publication years and author names. To identify illnesses, therapies, and results, healthcare professionals examine medical records. The benefits of neural network-based IE make it an essential technology, even in the face of obstacles like high processing demands, a need for a lot of training data, and ambiguity in language. In an era of information overload, it increases accuracy, saves time, and makes knowledge more accessible.

1. Information Extraction Using Neural Networks By Dibosh Biswas Masters of Library and Information Science Jadavpur University Roll no: 11 [email protected]

2. Information Extraction (IE) 1 Neural Network 2 Neural Network:- A neural network is a system of algorithms that tries to recognize patterns by simulating the way neurons in the brain send signals. I nformation Extraction (IE):- Information extraction is the task of pulling out specific pieces of information (like names, dates, places, relationships, events) from large collections of unorganized data. I nformation Extraction Using Neural Networks Topics Covered

3. INPUT DOCUMENT ↓ Sentence Analysis ↓ Assigning Deep case ↓ Network creation ↓ Question Analysis ↓ Search in Neural Network ↓ Back track to knowledge ↓ OUTPUT ANSWER Information Extraction Using Neural Networks

4. Information Extraction Using Neural Networks Step 1: Take input (text data) from the user. Let say document contains the following details “Dr. S.R. Ranganathan was born in Tamil Nadu on 12 August 1892”

5. Information Extraction Using Neural Networks Step 2: Divide the input sentences into words and give unique id each word ID Words Id1 Dr. S.R. Ranganathan Id2 Born Id3 Tamil Nadu Id4 12 Aug 1892

6. Information Extraction Using Neural Networks Step 3 : Extract Knowledge from input data Knowledge ID Knowledge K1 Dr. S.R. Ranganathan was born in Tamil Nadu K2 Dr. S.R. Ranganathan was born on 12 August 1892

7. Information Extraction Using Neural Networks Step 4 : Word type to each word ID Words Word type Id1 Dr. S.R. Ranganathan who Id2 Born what Id3 Tamil Nadu where Id4 12 Aug 1892 when

8. Information Extraction Using Neural Networks Step 5: Assign deep cases to each word ID Words Deep Case Id1 Dr. S.R. Ranganathan Agent Id2 Born Application Id3 Tamil Nadu Place Id4 12 Aug 1892 Date

9. Information Extraction Using Neural Networks Step 6: Relationship between words and knowledge First Word Second Word Knowledge ID Id1 Id2 K1 Id2 Id3 K1 Id1 Id3 K1 Id1 Id2 K2 Id2 Id4 K2 Id1 Id4 K2 ID Words Id1 Dr. S.R. Ranganathan Id2 Born Id3 Tamil Nadu Id4 12 Aug 1892 Knowledge ID Knowledge K1 Dr. S.R. Ranganathan was born in Tamil Nadu K2 Dr. S.R. Ranganathan was born on 12 August 1892

10. Information Extraction Using Neural Networks Step 7 : Create the netw ork based on the sentence Dr. S.R. Ranganathan was born in Tamil Nadu on 12 August 1892 Dr. S.R. Ranganathan was born in Tamil Nadu Dr. S.R. Ranganathan was born on 12 August 1892 Agent Dr. S.R. Ranganathan Action Born Location Tamil Nadu Time 12 August 1892

11. Nee d to analyse each word from the question Information Extraction Using Neural Networks Step 8: Neural netw ork created “ Where Dr. S.R. Ranganathan born?” Words Dr. S.R. Ranganathan Born Where

12. Information Extraction Using Neural Networks Step 9: Search the words from the netw ork and extract the knowledge Here K1 and K2 which relates both the words ‘ Dr. S.R. Ranganathan ’ and ‘ born ’. K1 answers “where” whereas K2 answers “when” . First Word Second Word Knowledge ID Id1 Id2 K1 Id2 Id3 K1 Id1 Id3 K1 Id1 Id2 K2 Id2 Id4 K2 Id1 Id4 K2

13. Step 10: Required knowl edge Information Extraction Using Neural Networks Output the selected knowledge unit K1 Dr. S.R. Ranganathan was born in Tamil Nadu

14. Information Extraction Using Neural Networks

15. Thank you! Thank you for your attention. I hope this presentation helped you understand the Information Extraction process using Neural Networks.