Nigerian economic intelligence
that banks can actually cite.

What Asotele solves

Nigerian economic decisions get made every day — treasury desks setting FX exposure, ministers announcing policy, businesses pricing loans, journalists writing about the economy. The data those decisions need exists, but it’s scattered across CBN circulars, NBS PDFs, news reports, tax gazettes, and Excel files nobody can find. Getting a straight answer with a citation takes hours. Most AI chatbots just make numbers up.

Asotele fuses those sources into one system and puts guardrails around the answer — every figure cited to its original source, refusal when the data isn’t there, and a visible correction the same day if we get it wrong.

Who this is for

What we deliberately don’t do

• Predict tomorrow’s FX rate. No one can — asking is a category error. We report what happened; scenario tools show conditional projections.
• Give investment advice. That’s the fiduciary duty of the desk making the trade. We give the data.
• Fill data gaps with plausible-sounding guesses. If we don’t have it, we say we don’t have it — that’s the whole point.

What’s public, what’s paid

Ambiguity here would be unfair to both audiences. Here’s the split, deliberately explicit:

The paid tier (banks, asset managers) subsidises the free tier. If citable Nigerian economic data required a Bloomberg-scale annual subscription, most Nigerian citizens, SMEs, journalists, and researchers would be locked out. That’s the public-good frame.

Recent project updates

• 2026-06-10
  The Pidgin corpus is built. The number didn't move. Here's what we learned.
  Following the honest 0.460 macro-F1 reported in the 2026-06-08 Pidgin post, we built the BBC News Pidgin domain corpus the LINGUA Africa proposal scopes — 776 articles, 41,208 cleaned paragraphs, four years of coverage. We then ran the workstream-2 hypothesis end-to-end: AfriBERTa masked-LM continued pretraining on that corpus, then re-fine-tune the sentiment classifier, then test on AfriSenti. Result: +0.0009 macro-F1. Functionally zero. Adding class-weighted loss on top: +0.0052 best-case. The bottleneck isn't Pidgin representation — it's the structural 1.4% neutral class in AfriSenti pcm, and no amount of Pidgin domain text will move that wall. This is the pivot post: the right workstream-2 measurement target is a hand-labelled BBC Pidgin sentiment test set, not the AfriSenti held-out test. We've built the sample; the labels are next.
• 2026-06-08
  Pidgin closes the loop — and the honest F1 gap that comes with it
  The fourth and final AsoteleLingua language pilot — Nigerian Pidgin (pcm) — fine-tuned overnight on AfriSenti SemEval-2023 and lands at test macro-F1 0.460. That's substantially below the Hausa 0.779, Igbo 0.782, and Yoruba 0.715 results from earlier in the week. The 4-of-4 milestone is real; so is the gap. This post lays out why Pidgin lags, what it means for the workstream-2 evidence in the LINGUA Africa proposal, and the grounded-knowledge layer that landed alongside it.
• 2026-06-04
  Yoruba lands, Igbo runs tonight — and the Igbo finding the proposal didn't expect
  AsoteleLingua now covers two of its four target Nigerian languages. The Yoruba pilot — fine-tuned AfriBERTa-Large on AfriSenti SemEval-2023 Yoruba plus a precision-tuned BBC Yorùbá financial scraper — lands at a +0.216 confidence lift over the multilingual baseline, replicating the +0.253 Hausa pattern from earlier this week. Igbo training kicked off tonight; the surprise was discovering, while scoping Igbo source coverage, that Igbo-language online finance content is structurally near-absent — itself the strongest possible argument for the LINGUA Africa workstream.

All updates →