The OpportunityFinder® algorithm has now exceeded 50,000 terms in its lexicon for detecting petroleum systems automatically in text. This is combined with hundreds of thousands of labelled data for machine learning. These can support laser like tasks, improve search & discovery, insights, knowledge mining and also support the tuning of very large language models.
We are using Deep Learning to leverage the unique 45,000 petroleum system related textual clues in OpportunityFinder®.
Designed for automation, the clues combined with auto-annotation of millions of sentences allow a deep learning model to generalise (learn). This enables the detection of valid clues in geoscience text (reports, presentations, papers) not present in the original lexicon.
Whilst an algorithm will never read text like a Geoscientist, we can teach it some elements. The advantage is reading differently to us, and processing larger volumes than any person can read.
This allows us to detect evidence, join the dots and stimulate business ideas we may not have had without the assistance of the algorithm.
OpportunityFinder® is being tested within a renewables geothermal project in collaboration with the British Geological Survey. BGS are investigating mine water in underground abandoned coal mines as a low carbon sustainable heat source for housing and manufacturing, and have several other potential use cases for knowledge extraction from their data archives to meet the challenges of decarbonisation and resource management.
#geothermal #renewables #naturallanguageprocessing #artificialintelligence #geoscience #energy #bgs
Announcing V2.0 released. Discover new oil & gas exploration ideas, leads, plays and opportunities in your unstructured text. Also exploit your unstructured text for analogues.
OpportunityFinder® – “first of its kind” pattern based geoscience search.
Example showing autoclassification output from GeoClassifier® from a selection of public domain geoscience documents. The proportion of topics are clustered in a Pearson dendrogram heatmap. Those above the mean are in red, below the mean in dark blue relative to the corpus/collection. Easy to see clusters of documents predominantly about certain topics and to spot ‘anomalies’ – which can be interesting to see and read further.
A new class has been added to the text GEOCLASSIFIER Algorithm
It can now predict sentences, paragraphs, pages or documents related to Palaeogeography and automatically tag them.
Image Credit: Map of the Late Cretaceous
To welcome in 2021 we are gifting GEOCLASSIFIER® – a geological machine learnt text classifier to not-for-profit organisations.
This assists information searching, filtering and discovery of geoscience topics in text. Even documents predominantly about one topic, often reference other geoscience topics buried within their pages. Automatically surfacing these topics could lead to insights that may otherwise go unnoticed.
Over 125 Million words from public geological texts were used to build the models.
The models in GEOCLASSIFIER® enable the automatic classification of text by industry sector; Metals and Mining, Engineering, Environmental, Geothermal, Hydrogeology, Petroleum and Planetary Geology.
They also classify by topic including; Mineralogy, Petrology, Sedimentology, Igneous, Metamorphic, Lithology, Volcanology, Commercial, Palaeontology, Geophysics, Tectonics, Geochemistry, Diagenesis, Hydrothermal, Glaciology, Geomorphology and Stratigraphy.
#artificialintelligence #machinelearning #textmining #geology #cognitiveassistant
The OpportunityFinder® algorithm now contains over 40,000 terms & phrases engineered specifically to be used to automatically extract geoscience knowledge in text for exploration.
Infoscience Technologies is 2 years old this month! To celebrate, here are some Exploration Geoscience KnowledgeGraphs automatically extracted from public domain geoscience content (documents, papers and presentations) by the OpportunityFinderⓇ algorithm.