Technology
Build the best solutions using the most relevant technology
Abeyon will implement innovative artificial intelligence (AI) & machine learning (ML)-based models using Natural Language Processing (NLP) to analyze unstructured data, derive knowledge and insights and provide an efficient way to search and retrieve data.
Abeyon will develop personalized solution sets using the best and most relevant AI technologies in NLP based on extensive research and review of the data provided by the company’s expert team. Solution sets will be built using open source tools to enable security, visibility into the technology, continuous improvements and the ability to deploy the solution in any infrastructure (local, private or public cloud).
Abeyon will review, analyze, extract, parse, classify and index text from large unstructured and structured data corpus using a set of tools collectively known as Smart Knowledge Extraction (SKE).
Information Extraction
Key phrases and relation tuples are the basic knowledge units that together represent the semantic meaning of a document, and generally represent an indicator of the potential “uniqueness” of a document. Open IE unlike traditional information extractions that focus on generating well-defined requests over predefined target relations on homogeneous datasets, is designed to extract all types of relations from text and scales to large, heterogeneous datasets. The Transformer based Neural Network significantly outperforms other approaches by learning the sequence-to-sequence task directly.
Another important technology within Information Extraction (IE) is Sentence Extraction. Sentence Extraction locates the most relevant sentences in documents both abstractive and extractive. Powerful sentence extraction models are able to select valuable, salient, and representative sentences from documents with variant-lengths, from short breaking news to pages-long official papers, manuals or statements by reflecting core semantic meanings of the source document, relationships between sentences, and knowledge from human annotation. Sentence Extraction is delivered through the implementation of attentive encoder-based models and through methods of reinforced training.
Named Entity Recognition
Auto keyphrase extraction tools automatically identify and prioritize key phrases from a given body of knowledge without human labeling effort and with limited, shallow linguistic analysis. The output from keyphrase extraction is a ranked list of phrases with decreasing quality. This can be used to display the most relevant results to the researcher. Key phrases can also be used to categorize and classify results to the most relevant topic. These key phrases provide quick visibility into the subject matter of the research.
Keyphrase Extraction
Key phrases and relation tuples are the basic knowledge units that together represent the semantic meaning of a document, and generally represent an indicator of the potential “uniqueness” of a document. Open IE unlike traditional information extractions that focus on generating well-defined requests over predefined target relations on homogeneous datasets, is designed to extract all types of relations from text and scales to large, heterogeneous datasets. The Transformer based Neural Network significantly outperforms other approaches by learning the sequence-to-sequence task directly.
Natural Language Search
When dealing with large amounts of unstructured data and looking to gain insights and meaning from these data, a simple Boolean search does not provide the extensive search capabilities necessary to review complex relations between topics, issues, new terms and languages. The search needs to go beyond simple keyword matching. An effective search needs to include indexed metadata from Smart Knowledge Extraction, related documents findability, external information identification, document similarity information and support for rich queries and semantic search. Abeyon will implement powerful natural language search capability where user can search using natural language (Ex: Show me sales in the past 6 months in the North east). Say bye to multiple search criteria pull downs, search faster, better and easier using NLS.
Read our insights story on NLS.