BM25 scores
From Dontopedia, the open, paraconsistent wiki. (Last updated 2026-06-09.)
BM25 scores has 9 facts recorded in Dontopedia across 5 references, with 2 live disagreements.
Mostly:rdf:type(3), computed by(2), used by(1)
Maturity scale
raw canonical shape-checked rule-derived certifiedInbound mentions (9)
Other subjects in dontopedia point AT this entity as a value. These are inverse relationships — e.g. "X motherOf this subject" — and answer questions the forward facts can't. Grouped by predicate.
combinesCombines(2)
- Score Combination
ex:score-combination - Weighted Sum
ex:weighted-sum
computedFromComputed From(2)
- Combined Scores
ex:combined-scores - Hybrid Score
ex:hybrid-score
returnsReturns(2)
- Bm25 Retrieval Function
ex:bm25-retrieval-function - Bm25 Retrieval Function
ex:bm25-retrieval-function
controlsWeightOfControls Weight of(1)
- Alpha
ex:alpha
hasComponentHas Component(1)
- Retrieval System
ex:retrieval-system
usesScoringMethodUses Scoring Method(1)
- Sparse Retrieval Microservice
ex:sparse-retrieval-microservice
Other facts (8)
The long tail: predicates that appear too rarely to warrant their own section. Filter or scroll to find a specific one. Each row links to its source.
| Predicate | Value | Ref |
|---|---|---|
| Rdf:type | Output Array | [1] |
| Rdf:type | Score | [2] |
| Rdf:type | Metric | [4] |
| Computed by | Bm25 Retrieval | [3] |
| Computed by | Linear Kernel | [3] |
| Used by | Sparse Retrieval Microservice | [4] |
| Computed for | Test Document | [5] |
| Compared Against | Training Documents | [5] |
Timeline
Timeline axis is valid_time — when each source says the fact was true in the world, not when Dontopedia learned about it. Retracted rows are kept for provenance; coloured stripes indicate the context kind.
References (5)
ctx:claims/beam/8036737b-9c5e-4cf6-8fd5-40137132613b- full textbeam-chunktext/plain1 KB
doc:beam/8036737b-9c5e-4cf6-8fd5-40137132613bShow excerpt
Finally, you can combine the results from both sparse and dense retrievals. One common approach is to use a weighted sum of the scores from both methods. Here's a more complete example: ```python import numpy as np from sklearn.feature_ex…
ctx:claims/beam/b0390377-17cd-4838-999f-26ca02c6c6a4- full textbeam-chunktext/plain963 B
doc:beam/b0390377-17cd-4838-999f-26ca02c6c6a4Show excerpt
- We use a pre-trained BERT model to generate embeddings for documents and the query. - `cosine_similarity` computes the similarity between the query embedding and document embeddings. 3. **Combining Scores**: - We combine the BM2…
ctx:claims/beam/07b00e3a-dd0e-40bb-a9be-bbdf1ac254da- full textbeam-chunktext/plain1 KB
doc:beam/07b00e3a-dd0e-40bb-a9be-bbdf1ac254daShow excerpt
with torch.no_grad(): doc_outputs = model(**doc_inputs) query_outputs = model(**query_inputs) doc_embeddings = doc_outputs.last_hidden_state.mean(dim=1) query_embedding = query_outputs.last_hidden_state.mean(dim…
ctx:claims/beam/961aaaa1-3f78-41a4-b639-fb057c9f07c8- full textbeam-chunktext/plain1 KB
doc:beam/961aaaa1-3f78-41a4-b639-fb057c9f07c8Show excerpt
4. **Final Ranking**: Rank the combined results and return the top-k documents. ### Step 2: Architectural Components To achieve 2,000 queries/sec with 99.9% uptime, you need to design a scalable and fault-tolerant architecture. Here are t…
ctx:claims/beam/9669963d-f7d7-452d-a9ec-0cf09ed6be1d- full textbeam-chunktext/plain1 KB
doc:beam/9669963d-f7d7-452d-a9ec-0cf09ed6be1dShow excerpt
predictions.append(predicted_label) return predictions # Make predictions predictions = predict_labels(test_df, bm25, train_df) # Calculate the recall score recall = recall_score(test_df['label'], predictions, average='binary'…
See also
Keep researching
Missing something or suspicious of what's here? Kick off a research session — a Claude agent will investigate, cite its sources, and file new facts into a dedicated context you can review before accepting into the shared view.