Add lectures 9

NLP/lectures9.md

@@ -0,0 +1,187 @@
+# Lectures 9
+
+## Word senses and disambiguation
+
+### Lemma and Wordform
+
+- _Lemma_: base form of a word
+- _Wordform_: inflected form of a word
+
+```
+Wordform    lemma
+banks       bank
+sung        sing
+plese       plesati
+```
+
+Lemmas have senses, One lemma "bank" can have many meaning.
+
+S1: "a bank can hold the investments in a custodial account"
+S2: "as agriculture burgeons on the east bank the river will shrink even more"
+
+### Homonys (enakozvocnice)
+
+Words that share a form but have unrelated, distinct meanings.
+
+```
+bank1: financial institution, bank2 : sloping land
+bat1: club for hitting a ball, bat2: nocturnal flying mammal
+bass: stringed insturment, bass: fish
+```
+
+### Polysemy (večpomenskost)
+
+How do we know whan a word has more than one sense?
+
+LLM: Are the senses of word "severe" in S1 and S2 identical?
+
+- zero shot
+- few shot
+- fine tuning
+
+S1, S2, <word> SAME/DIFFERENT
+
+### Synonims
+
+Word that have the same meaning in some or all contexts.
+
+– filbert / hazelnut
+– couch / sofa
+– big / large
+– automobile / car
+– vomit / throw up
+– Water / H20
+
+### Antonyms
+
+Senses that are opposites with respect to one feature of meaning.
+
+### Hyponyms
+
+House is a hyponym of building, building is a hyponym of property.
+Car is hyponym of vehicle.
+Mango is a hyponym of fruit.
+
+A sense A is a hyponym of sense B if being an A entails being a B.
+
+### Meronymy
+
+A leg is a part of a chair; a wheel is a part of a car.
+
+Wheel is a meronym of car and car is holonym of wheel
+
+### Supersenses
+
+A words supersense can be useful coarse-grained representation of its meaning.
+
+```
+fasttext } merge all sense into one vector
+word2vec }
+```
+
+BERT - one vector per context.
+
+### Path based similarity
+
+Two concetps are similar if they are near to eachoter in thesaurs hiearchy. - Have a short path between them - concepts have path 1 to themselves
+
+```
+pathlen(c1, c2) = 1 + number of edges in the shortest path between c1 and c2
+
+sim_path(c1, c2) = 1/pathlen(c1, c2)
+
+wordsim(c1, c2) = max simpath(c1, c2) for all senses of c1 and c2
+```
+
+#### Path Diagram
+
+- **medium of exchange**
+  - **currency**
+    - **coinage**
+      - **coin**
+        - **nickel**
+        - **dime**
+    - **fund**
+      - **budget**
+  - **money**
+    - **standard**
+      - **scale**
+        - **Richter scale**
+
+### Calculated Similarities
+
+- `simpath(nickel, coin) = 1/2 = 5`
+- `simpath(fund, budget) = 1/2 = 5`
+- `simpath(nickel, currency) = 1/4 = 25`
+- `simpath(nickel, money) = 1/6 = 17`
+- `simpath(coinage, Richter scale) = 1/6 = 17`
+
+### Dekang similarity
+
+This year skipped...
+
+### Evaluating similarity
+
+Extrinsic evaluation: - Question answering - Spell checking - Essay grading
+
+Intrinsic evaluation: - Correlation between algorithm and human word similarity raitings - taking TOEFL multiple-choice vocabulary test
+
+### Word sense disambiguation (WSD)
+
+Given a word context and a fixed inventory of potential senses, decide which sense is intended.
+
+### Word in context (WiC)
+
+Word in Contex (WiC) dataset: determine if two sentences contain a word with the same or different sense, Contains senses mostly from the WordNet.
+
+### WiC fro WSD evaluation
+
+    - WSD on S1 and S2; if the same sense is predicted, the answer is True, otherwise False.
+
+### WiC for WSD prediction
+
+    - WiC classifier; apply S and to sense glosses, predict the sense with the highest score.
+
+What is a classification ceiling for WiC? around 80%.
+
+### Word sense induction
+
+How to use WSD to detect new senses?
+
+```
+bass
+1 ------ G1
+2 ------ G2
+3 ------ G3
+....
+8 ------ G8
+9 ------ G9 <- new sense
+```
+
+Compare new sense with all the existing senses. If the similarity is low, then it is a new sense.
+Clustering fails in practice, because of the huge overlap of the senses.
+
+## Affective computing
+
+Why compute affective meaning?
+
+- Sentiment towards politicians, products, countries, ideas
+- Frustration of callers to a help line
+- Stress in drivers or pilots
+- Depression and other medical conditions
+- Confusion in students talking to e-tutors.
+- Emotions in novels
+
+### Scherer's typology of affective states
+
+- **Emotions**: short-lived, intense, specific, directed at an object
+- **Moods**: longer-lived, less intense, not directed at an object
+- **Interpersonal stances**: attitudes towards others
+- **Attitudes**: general evaluations of objects
+- **Personality traits**: stable individual differences
+
+### Emotions
+
+Atomic basic emotions: - A finite list of 6 or 8. from which all other emotions are constructed.
+
+Dimensions of emotion: - Valence: positive or negative - Arousal: strong or weak - Dominance: in control, active vs controlled, passive

NLP/main.md

@@ -7,3 +7,4 @@
 - [Lectures 6](https://github.com/spagnoloG/FRI-ZAPISKI/blob/main/NLP/lectures6.md)
 - [Lectures 7](https://github.com/spagnoloG/FRI-ZAPISKI/blob/main/NLP/lectures7.md)
 - [Lectures 8](https://github.com/spagnoloG/FRI-ZAPISKI/blob/main/NLP/lectures8.md)
+- [Lectures 9](https://github.com/spagnoloG/FRI-ZAPISKI/blob/main/NLP/lectures9.md)