Patent Run | before 2022Q2 Presentation
It’s been a long time since last update for this project but here it is!
After a quick re-run of the code – and fixing some inevitable bugs which under-development code has – I saw that the “stats” deviated quite a bit from the 29 May run which was used for the last Patent-Time Plot. This was due to the fact that the scoring system was under heavy work and I spent way to much time reading patents to fine-tune this document causing it to “harden” after mentioned run.
The Patent-Time plot uses the tuv_G2.csv files. After being a bit shocked to see the reduction from 125 runs I got an equivalent of only 95 with nearly identical search parameters. Looking briefly into the differences is a golden opportunity to catch-up and potentially improve some stuff! For transparency files are also being shared publicly here.
Comparison of runs
The jobs compared was run 29 of May and 13 of August. The stats are quite different!
I would expect the reason be primarily due to differences in scoreKeyword.xlsx which provides weights and keywords for how each patent is scored. I could only find slight differences where some additional keywords was added to remove eyewear. Names of competitors was added to it would be flagged in the resulting .csv files – which should not affect scoring.
Next step is to compare the G2 groups for the two runs. Copy sorted patent IDs and use COUNTIF to see which patents from May was discarded in August run.
This is manually grueling work. Going into each patent and review scores and keywords – when all you want to make is plots is painful, but someone has to do it! The 4th patent was a Vivo patent that refers to T-lens and has a score of 147. Now I can look into crawlerALL.csv for the August run to find out that it only got 95 points. Seeing how the three first patents was removed correctly, we’ll update scoreKeywords.xlsx to capture T-lens for the next run! These weights will never be perfect, but they got slightly better today!
Latest Run
From the results we can see a few things. The O-FILM subsidiary Jingxi Jinghao has a large amount of TLens patents and we know that they offer a AR Main Camera with TLens today. Their work with TLens is continued. We can also see a ramp-up in patents from Sunny and Truly-Opto.
Some more quality checking is required before analyses of players is due! The translation dictionary also needs a polish!
