MarketPsych data, available via MCP

Sentiment, emotions, topics and events extracted from earnings and other corporate conference calls. Every sentence of every call is scored, and the results aggregate by document, section and speaker, so prepared remarks can be separated from Q&A, and management from the analysts questioning them. Full transcript text is available alongside the scores.

Numeric sentiment, emotion and thematic time series, one series per asset, aggregated from thousands of news and social media sources. More than a hundred indices per asset class, from buzz and sentiment through to innovation, management trust, layoffs and litigation. Minutely, hourly and daily frequencies, with coarser windows aggregated on request.

The outside-in view of corporate conduct: what news and social media report about a company, rather than what the company reports about itself. Two layers are available. Core gives 1–100 percentile ranks across three pillars and ten categories, computed within a company’s TRBC industry using the same weighting as LSEG ESG, so the self-reported and media views sit side by side. Advanced gives the granular counters underneath, roughly a hundred per company, including controversy topics such as tax fraud, industrial accident, child labour and anti-competitive acts.

A single daily 1–100 percentile rank for each stock, distilled from LMA sentiment and fitted as a returns model in the StarMine family. Three component ranks (Equity, Business and Management) break the headline score into its parts, and six regional models respect local differences in how media coverage moves prices. The signal is close to uncorrelated with the usual factors, which is the case for running it alongside fundamental models rather than in place of them.

Emerging narrative clusters, detected daily across global news and social media. Each day’s buzz for every extracted term is tested against its own hundred-day baseline; the surges become themes, which are then named, summarised and categorised, and scored for market relevance and urgency. Every theme carries the entities it touches, each with its own buzz and sentiment, so a narrative can be traced straight to the assets it moves.

Sentence-level content, covering news, social media, press releases and filings, where every sentence carries its own sentiment and its own tagged entities. Where LMA gives you the number, Radar gives you the sentences behind it. It is the tool for discovery work: building a thematic basket from what is actually being written, mapping which companies a driver hits hardest, or reading the evidence behind a score.

The NLP engine behind every dataset above, pointed at text that you supply: research notes, filings, broker commentary, internal reports. It returns per-sentence sentiment, linked entities and keyphrases, and resolves company mentions to PermIDs so its output joins directly to LSEG and MarketPsych data. The same engine powers Yahoo! Finance and the iPhone Stocks app, and is available as NLP-as-a-Service for generative AI, alpha generation and entity recognition.

MCP access uses your existing MarketPsych or LSEG credentials, works with any MCP client, Claude and ChatGPT among them, and sits alongside the delivery channels you may already use: API, SFTP, Snowflake, Databricks and the MarketPsych web app. Existing clients can have MCP enabled on their current entitlements.

New to the data? The MarketPsych Research Accelerator provides twelve months of access to the full LSEG MarketPsych suite (LMA, ESG, MMS and MTA) with supporting materials and quarterly updates, at nominal cost. It is built for teams that want to test signals across a complete research cycle.

To arrange a trial or discuss access, contact your LSEG representative or email [email protected].