3D Word Vector Plots

3D Word Vector Visualisations: The provision of free web tools for all geoscientists to easily explore hidden semantic relations in textual content may increase the chances of abductive discovery in our discipline.

Following on from previous posts on word vectors in space and time, in this example the ‘first’ axis is Lacustrine, ‘second’ is Evaporite and ‘third’ is Pegmatite. The closer to 1, the more similar the geological age is to each axis. I chose this to show obvious differences (see Precambrian for example in orange).

It is possible to choose the main entity such as Lithostratigraphy, Minerals, Well/Boreholes, Prospects, Deposits, Projects, Reservoirs etc, then query on whatever you wish to explore. Combining statistical Text Embeddings with Natural Language Processing and Knowledge Engineering. The speed and interactivity of such plots can be engaging – removing any barriers between the geoscientist and the information corpus allowing them to choose whatever queries they wish.

Many discoveries are serendipitous, perhaps we can further ‘nudge’ or ‘facilitate serendipity’ by using such approaches..

There are existing Text and Data Mining services and tools already available in the geosciences (GeoDeepDive xDD and Deep-time Digital Earth). These are very useful, although not every geoscientist can easily freely use them or apply to the content they wish to target. There may be a gap in which to add additional lightweight easy to use Digital Assistants to complement digital capabilities for the geoscience discipline.

#naturallanguageprocessing #unstructureddata #geoscience #geology #digital #subsurface #earthscience #artificialintelligence