- Incrementality in Compositional Distributional Semantics (SFB 732 D10, 2014-2018):
Project D10, SFB-732, German National Science Foundation (DFG) (2014-2018). Role: Researcher
The principle of compositionality is fundamental to model-theoretic semantics in the tradition of (Montague, 1973). Distributional models of lexical semantics are starting to account for compositionality as well: Novel syntax-semantics interfaces map syntactic categories onto algebraic types -- e.g. nouns to vectors, adjectives to matrices, etc. -- and define general operations to combine objects of these types (Baroni & Zamparelli, 2010; Grefenstette et al., 2013). This approach offers an attractive framework for distributional semantics construction, but research has so far been limited.
The goal of Project D10 is to contribute to research on tensor-based compositional distributional semantic models by developing a syntax-semantics interface with three properties:
(a) it will be dependency-based rather than based on constituents;
(b) it will be incremental, that is, construct semantics in a left-to-right manner;
(c) it will incorporate a notion of plausibility for (partial) analyses based on expectations at the level of individual composition operations.
The first property is important to develop syntax-semantics interfaces for languages with a more free word order. The second and third are well-known properties of human sentence processing
Project LOVe: Linking Objects to Vectors in distributional semantics - an EU, Marie Skłodowska-Curie project
EU,Marie Skłodowska-Curie project Nr.655577,H2020-MSCA-IF-2014 (2015). Role: Researcher
The project aims at building a framework to anchor corpus-based meaning representations to the external world. I was involved in extracting and establishing the entity-attribute relationships encoded in distributed embeddings of Named Entities.
- Natural Language Interfaces to Databases
The aim of this project was to create a robust and easy-to-customize Natural Language Interface to Database (NLIDB) which can be quickly deployed by any organisation which uses a database to store information (a university, company, small-scale business, etc). An NLIDB is a software which takes a query from a user in natural language (say English) and extracts relevant answers from a database. This software removes the need for a person to have an in-depth knowledge of backend softwares (like Oracle, MySQL, etc) to get the required information.
- Online Character Recognition:
The aim of the project was to extract a set of feature vectors for the data given such that once a classifier is trained on those feature vectors the accuracy on the data given is the highest possible that can be achieved.
- Creating a Neural Network Classifier:
The project involved creating a Neural Network Classifier working on a Feedback mechanism given a set of classes with a specified feature set. Classification is performed once the classifier is trained using a sample training data-set.
- Identification of Tense, Aspect and Modality in text to improve Machine Translation accuracy:
The project delt with identification of Tense, Aspect and Modality(TAM) in a sentence of given text for the purpose of machine translation. Their identification in source text lead to improved translation of verbs and TAMs in the target language, making the translation more human readable.
- Handing gerunds in English for Machine Translation:
The project involved the use of basic NLP techniques. Gerunds as a word category are mostly incorrectly identified during translation and the project's aim was to identify the occurrences of gerunds in English in various contexts. Then translation is done according to the context implied in the sentence with correct TAMs.
- Effect of Discourse Connectives in Languages:
The goal was to understand the role played by discourse connectives in a language which helps in establishing a meaningful dialogue. Correct interpretation of the context and scope of discourse connectives leads to a higher accuracy during machine translation.
- Study of Computer Simulation of Ashtadhyayi - Some Insights:
(By Pawan Goyal, Amba Kulkarni and Laxmidhar Behera)
The purpose of this project was the indepth understanding of how Paninian Grammar can be implemented computationally to most of the free-word ordered as well as ordered languages. Parsers based on Computational Paninian Grammar not only give a syntactic dependency parse but also provide sufficient semantic information giving an edge to machine based translations.
- Website: ILMT Module Management:
The ILMT module management portal facilitates the maintainance and management of applications, tools and resources that have been created for 9 Indian language pairs i.e. 18 MT systems with the help of a consortium of 9 institutions. The website provides for uploading, version management, testing and verification processes which are essential to the project. All the website functionalities are provided with fully automated initimation and feedback mechanism. The Indian Languages Machine Translation project ('SAMPARK') is a web-based machine translation system which will produce machine translation (MT) from one Indian language (IL) to another Indian language.
Role: Developer, site administrator and manager
- Website: International Conference on Natural Language Processing (ICON) & NLP Tool Contest Portal:
ICON is one of the key conferences for the research community of the Indian subcontinent to recognize the work done on Indian languages. I have managed the 1) Main conference website; 2) Paper submission and review website and; 3) NLP tools contest website, since 2008.
Role: Developer and site manager
- Grievance System for Jaipur Municipal Co-corporation (Government E-governance Project):
The System is a desktop – web based integrated application. It is being used to record customer complaints. The online entries by the customers are recorded by operators at the municipal office and directly entered by the customer online. The role of operators is maintained and controlled by administrators of the system. The scope of the complaints is automatically determined by the area as well as the role of the operators.
Role: Implementation, Black-Box testing and part of the maintenance team of the entire system.
- House-Tax System for Jaipur Municipal Co-corporation (Government E-governance Project):
The System is a desktop application designed for a chain of gymnasiums. It maintains the records of the customers taking membership. Gives an overall performance based on the balance sheets, reports on various criteria and a billing system.
Role: User Interface, Class design and coding.
- Product Catalog System:
The System was developed to provide an easy display for the wide range of products developed by the organizations during exhibitions and take on-the-spot orders from customers specially during the international events. The desktop application when connected to the parent application updates the new orders acquired into the main application and hence helps in the integration of data and orders collected from various locations and events into the main application maintained at the head office.
Role: System design, implementation.