Investment Intelligence

Boost your competitive investment insight by combining proprietary information and global market data and further refining it.

No single database can offer good company profiles and transaction coverage
Industry and product classifications in databases are scarce and often inappropriate
Events and company profile updates should be extracted from text, e.g. news

How You’ll Benefit

Speed up research, get profound company profiles and perform more precise target qualification.

Enhance offerings by adding new types of information and improving data quality.

Get better view on market and competitor transactions to tune up their corporate development strategies.

Integrate data from multiple vendors such as Factset, BvD, Refinitive, Crunchbase and Capital IQ.
Interlink and fuse data across proprietary and external databases with millions of records.
Reconcile and refine the data to provide appropriate links and classifications, e.g., industry sectors and technology fields.

Provide discovery, reporting and visualization facilities of unmatched sophistication in tracing multi-hop relationships.
Apply analytics for various tasks, e.g., semantic search for companies or suggestion of alternative investment targets.
Keep the graph up to date with updates from multiple structured sources and news and maintain data quality.

Integrate data from diverse sources into a single database.
Normalize and reconcile locations, industry sectors and other “strings to things”, i.e., references and unified master data structure: ontology, taxonomy or nomenclatures.
Match descriptions of one and the same entity (e.g., company) across different sources and fuse them in a single rich coherent entity profile interlinked to other entities in a knowledge graph. Cohort of machine learning techniques is used to go beyond the limitations of hand-crafted heuristics and extensive manual processing.

The task: Get adequate classifications by industry sectors, technology fields, business model or other criteria.
The challenge: Even the most robust company databases include industry sectors only for the top 1% of the biggest companies on Earth. Those classifications are often irrelevant for specific analytical purposes, e.g. GICS and ICB does not have sector Sport.
The solution: Reclassify companies automatically using a combination of machine learning techniques such as word embedding models (like GLOVE) and classification models (e.g., BERT) pre-trained on Wikipedia or a similar dataset with big volume of manual classifications against a very rich folksonomy of categories (industries, tech, etc.).

Extract more companies from news. Company data vendors do not offer good coverage of less developed markets, e.g., Vietnam or Malaysia. Business news from these markets covers plenty of mid-sized entities, which are missing in the databases.
Extract relationships and enrich company profiles regarding industries, products, business models, technologies, related events, officers and more. For this purpose the capabilities of Ontotext’s Platform to use big knowledge graphs for text analysis are augmented with tailor-made pipelines using BiLSTM neural networks

Data is managed in Ontotext’s GraphDB engine, which offers a powerful combination of structured queries (SPARQL), reasoning (inferring new relationships over transitive control chains, location and category hierarchies, etc.), graph analytics (e.g., entity importance via PageRank), geo-spatial constraints (e.g., near by), full-text search and semantic similarity, based on word and graph embedding. This allows for powerful faceted search interfaces, neat incremental suggestions and comprehensive pattern discovery based on multi-hop relationships.
Alternative investment targets can be detected using inductive logic programming methods like CN2 to analyze transactions, detect indirect associations and derive logical rules describing investment strategies

Continuously update company data graph with updates from the different sources. Efficient data processing and analytical pipelines guarantee that very large graphs of company data can be updated with substantial changes from multiple sources overnight.
Prevent data quality degradation after data updates. Comprehensive data quality analysis is performed to guarantee that updates do not jeopardize cleaning and reconciliations implemented initially. Continuously re-train entity matching models to improve accuracy over time.