MarketPsych's Comprehensive Suite of AI-Based Natural Language Processing Solutions
MarketPsych's seven products span the full range of AI-based natural language processing (NLP) solutions, from Tagging-as-a-service to data feeds (analytics) to tailored predictive models.
1. Scoring Engine
Tagging-as-a-Service platform for financial analytics, developed to process textual data from various sources.
2. Media Radar
Database of historical and live scores of text from global news and social media sources, for custom feeds and alerts.
3. Media Analytics
Time series of sentiment and topic scores. Data is delivered by asset and updated minutely, hourly, and daily.
4. ESG Media Analytics
Aggregated ESG topic and sentiment scores by company or country, updated minutely, hourly, and daily.
5. Transcript Analytics
Earnings conference call transcript tagging and sentiment analytics by sentence, speaker, section, and document.
6. Event Pulse
Automatic detection & summarization of key corporate events using generative AI. News summaries published in real-time.
7. Media Sentiment Model
Predictive model for global stocks built on Media Analytics scores and targeting 30-day returns.
Scoring Engine for Natural Text (SENT)
A Tagging-as-a-Service platform for financial text analytics
MarketPsych's Scoring Engine for Natural Text (SENT), is an Tagging-as-a-service platform for financial analytics. SENT was developed to process textual data from sources such as news, social media, fillings, transcripts, and brokerage research. Its output consists of dozens of attributes, some of the most important being:
- Topic classification:Over 1,000 hierarchical topics (e.g., accounting, emissions, litigation) and more than 4,000 events can be identified.
- Entity recognition:Tags references to millions of entities of over 20 types (companies, products, locations, commodities, people, and more).
- Sentiment analysis:AI-based classifiers tailored specifically for finance, ESG, and commodities contexts.
- Emotion profiling:Beyond sentiment, it identifies 14 emotions conveyed in the text, such as anger, fear, and optimism.
Figure 2: Schematic representation of the main attributes identified via SENT.
Clients use the output to generate standardized insights from their unstructured documents. SENT can be deployed on-premises, in the cloud, or via API. With SENT, the MarketPsych team is your outsourced NLP experts, keeping you updated on the latest advances.
MarketPsych Media Radar
Analyze themes being discussed in global media in real time.
MarketPsych feeds a collection of more than 300,000 global news and social media websites into SENT to generate a detail-rich feed with billions of data points. The API and User Interface (UI) access points allow filtering media information by various parameters, including free-text search, entity, and topic identifiers.
Clients use Media Radar to generate customized sentiment data feeds, create email alerts, track negative news about specific companies or asset classes, visualize public opinion on certain people (such as political candidates), get notified of the release of new products, and find companies associated with specific themes.
Figure 3: Media Radar on monitoring weight loss drugs
LSEG MarketPsych Analytics (LMA)
Media-based sentiment data feeds for creating predictive models.
For clients looking for a simplified data feed ready for quantitative analysis, we created the LSEG MarketPsych (Media) Analytics, LMA. This product is an aggregated sentiment feed derived from Media Radar.
It covers over 130,000 entities from asset classes such as companies, bonds, commodities, currencies, cryptos, stock indices, etc. The thousands of topics identified in Media Radar are condensed into a few dozen overarching indicators related to business and finance. LMA indicators go back as far as 1998 and are updated in near-real time (60-second updates).
Figure 4: Sentiment (blue) of news and social media posts referring to the Euro superimposed with the EUR/USD rate (black).
LSEG MarketPsych Analytics (LMA-ESG)
Media-based ESG data feeds for risk profiling.
While LMA aggregates business/finance-related topics, LMA-ESG aggregates Media Radar data into ESG-focused themes. Because companies and countries typically do not report their own negative ESG news, this feed excels in detecting greenwashing, controversies, and real-time ESG risks, thus revealing hidden ESG insights and providing a first-mover advantage.
Figure 4: In early 2020, a short-seller report about Luckin Coffee broke into the news. The ESG Controversies score (0 to 100, worst to best) fell accordingly. Two months later, an investigation found that the company had faked sales numbers
Typical clients are risk managers (controversy detection), ESG Funds, ESG ETF providers, ESG analysts (greenwashing detection), and consulting firms (ESG profiling).
MarketPsych Transcripts Analytics
Earnings call transcripts-based data feeds for research and modeling.
MarketPsych’s Transcripts Analytics, MTA, uses SENT to analyze the content of corporate calls including earnings conference calls. It covers over 15,000 companies with a history dating back to 2001. MTA allows for quickly filtering key themes, developing alpha models, and identifying investment risks. Users include quants, traders, research and ESG analysts, and risk managers.
Figure 5: Average financial sentiment of all sentences from Microsoft's quarterly earnings conference calls superimposed with its stock price
Event Pulse
Key corporate news summarized into an easy-to- digest feed.
MarketPsych’s AI-powered Event Pulse identifies and summarizes relevant global events for investors, advisors, and analysts. It uses the output of LMA to identify large increases in media attention about a company; then, a generative AI-powered engine outputs a headline and a summary of the event by analyzing the news that gave rise to the LMA spike. It also categorizes the event’s key topic and sentiment. Investors using Event Pulse streamline their workflows, learn the context for stock performance, and quickly identify opportunities and risks.
Figure 6: Event Pulse identifies increased media references to AMD in early 2023. It then summarizes and generates metadata based on the news responsible for the attention spike.
StarMine MarketPsych Media Sentiment Model (MMS)
Stock price prediction 30 days into the future.
The StarMine MarketPsych Media Sentiment Model, MMS, is a predictive model for global stocks based entirely on the indicators in LMA (thus entirely on the media information flow). Its output is an easy-to-use 1 to 100 score, which indicates the rank of a given stock according to its relative-to-other-stocks performance for the next 30 days. The model has held its training Sharpe ratio of 1 between top and bottom-ranked deciles for U.S. stocks since it was launched in January 2020. The signal is orthogonal to fundamental factors and is market cap agnostic.
Figure 6: Spread between U.S. stocks' top and bottom deciles according to the MMS model.
Active funds and traders use MMS as a stand-alone model or part of a suite of models for stock price forecasting.