Notes on first meeting » History » Version 16

Gyorgy Fazekas, 2012-02-23 12:54 PM

1 6 Gyorgy Fazekas
These notes are based on our initial meeting on 24 January 2012. The aim was to collect some use cases and have an initial idea on what needs to be done to extend or revise the existing Audio Features Ontology.
2 1 Gyorgy Fazekas
3 14 Gyorgy Fazekas
h1. Topics discussed 
4 1 Gyorgy Fazekas
5 14 Gyorgy Fazekas
A rough list of topics discussed during the first meeting:
6 14 Gyorgy Fazekas
7 1 Gyorgy Fazekas
** What are the main research use cases for an Audio Features Ontology (AF) ?
8 2 Gyorgy Fazekas
** Are they served well by the existing AF ? 
9 1 Gyorgy Fazekas
** If not, what are the most important extensions we need to do?
10 1 Gyorgy Fazekas
** Does the fundamental structure of the ontology need to be changed?
11 1 Gyorgy Fazekas
** What is the relation of AF to existing software, including:
12 10 Gyorgy Fazekas
13 8 Gyorgy Fazekas
 * software like: Sonic Annotator, Sonic Visualiser, SAWA, AudioDB other tools... 
14 1 Gyorgy Fazekas
 * and projects like: OMRAS2, EASAIER, SALAMI, new Semantic Media/Semantic Audio grants...
15 7 Gyorgy Fazekas
** Personal Objectives: what are we going to do with a modified/re-engineered ontology?
16 1 Gyorgy Fazekas
17 2 Gyorgy Fazekas
18 4 Gyorgy Fazekas
h1. Use cases:
19 1 Gyorgy Fazekas
20 12 Gyorgy Fazekas
Use cases discussed so far:
21 12 Gyorgy Fazekas
22 12 Gyorgy Fazekas
+Thomas:+
23 1 Gyorgy Fazekas
  
24 1 Gyorgy Fazekas
** drive audio effects -> adaptive effect (controlling effects)
25 1 Gyorgy Fazekas
** KM like use case: association of audio effects and audio features e.g. pitch shifter won’t change onsets
26 1 Gyorgy Fazekas
** part of the AFX ontology
27 1 Gyorgy Fazekas
** more audio features
28 1 Gyorgy Fazekas
** technical classification of audio effects
29 1 Gyorgy Fazekas
30 1 Gyorgy Fazekas
31 12 Gyorgy Fazekas
+Steve:+
32 1 Gyorgy Fazekas
** Finding structure, repeated sequences of features
33 1 Gyorgy Fazekas
** Beat related stuff, BPM (tempo, major/minor is it an audio feature, not necessarilty)
34 1 Gyorgy Fazekas
** Chords => Chord ontology
35 1 Gyorgy Fazekas
** Melody and notes
36 1 Gyorgy Fazekas
37 1 Gyorgy Fazekas
38 12 Gyorgy Fazekas
+George:+
39 1 Gyorgy Fazekas
** Improve SAWA
40 1 Gyorgy Fazekas
** Facilitate the development of intelligent music production systems
41 1 Gyorgy Fazekas
** Release large content based metadata repositories in RDF
42 1 Gyorgy Fazekas
** Re-release the MSD in RDF (??)
43 1 Gyorgy Fazekas
** Deploy a knowledge based environment for content-based audio analysis based on the concept of the Knowledge Machine that can combine multiple modalities
44 1 Gyorgy Fazekas
** Research reproducibility using Ontologies as a model to exchange research data.
45 1 Gyorgy Fazekas
46 6 Gyorgy Fazekas
47 1 Gyorgy Fazekas
h1. Open issues:
48 6 Gyorgy Fazekas
49 1 Gyorgy Fazekas
Some important questions to be decided on:
50 12 Gyorgy Fazekas
51 1 Gyorgy Fazekas
h2. Domain boundaries and scope:
52 1 Gyorgy Fazekas
53 14 Gyorgy Fazekas
** Are Musicological concepts outside the domain of an AF ?
54 12 Gyorgy Fazekas
** How about Physical features:
55 1 Gyorgy Fazekas
 
56 14 Gyorgy Fazekas
 * Acoustic features, 
57 14 Gyorgy Fazekas
 * Perceptual Features, 
58 14 Gyorgy Fazekas
 * DSP type feature, 
59 14 Gyorgy Fazekas
 * Musical Features (musically meaningful features related to acoustics) 
60 1 Gyorgy Fazekas
61 14 Gyorgy Fazekas
** The scope of the revised ontology may be:
62 1 Gyorgy Fazekas
63 14 Gyorgy Fazekas
 * Facilitate data-exchange for various purposes: (e.g. Linked Open Data, Research reproducibility, etc...)
64 14 Gyorgy Fazekas
 * Facilitate building intelligent/knowledge-based systems:
65 14 Gyorgy Fazekas
 ** How expressive the Ontology should be?
66 14 Gyorgy Fazekas
 ** What kind of reasoning services should be supported?
67 14 Gyorgy Fazekas
68 14 Gyorgy Fazekas
69 1 Gyorgy Fazekas
h2. Fundamental structure of the existing AF Ontology:
70 1 Gyorgy Fazekas
71 14 Gyorgy Fazekas
The Audio Features Ontology currently provides a core model which distinguishes between audio features based on two attributes:
72 14 Gyorgy Fazekas
73 14 Gyorgy Fazekas
# Temporal characteristics
74 14 Gyorgy Fazekas
# Data density
75 14 Gyorgy Fazekas
76 16 Gyorgy Fazekas
Alternative conceptualisations and some examples are summarised below:
77 1 Gyorgy Fazekas
78 1 Gyorgy Fazekas
!http://isophonics.net/sites/isophonics.net/files/FeatureConceptualisations.png!
79 16 Gyorgy Fazekas
*Fig 1.* Conceptualisations of content-based features.
80 16 Gyorgy Fazekas
81 16 Gyorgy Fazekas
The first dichotomy allows for describing features either instantaneous events (e.g. note onsets, tempo change), or features with a known time duration (notes, structural segments, harmonic segments, the extent of an STFT or Chromagram frame).
82 15 Gyorgy Fazekas
83 14 Gyorgy Fazekas
The second dichotomy addresses a representational issue, and allows for describing how a feature relates to the extent of an audio file: 
84 14 Gyorgy Fazekas
** whether it is scattered and irregularly occurs during the course of a track (i.e. sparse),
85 14 Gyorgy Fazekas
** or occurs regularly and have a fixed duration (i.e. dense).
86 13 Gyorgy Fazekas
87 11 Gyorgy Fazekas
!http://isophonics.net/sites/isophonics.net/files/AF_ontology_small.png!
88 11 Gyorgy Fazekas
89 9 Gyorgy Fazekas
90 6 Gyorgy Fazekas
91 6 Gyorgy Fazekas
The main scope of the ontology is to provide a framework for communication, feature representation, and describe the association of features and audio signals. Therefore it does not classify features, describe their interrelationships or their computation. With re- gards to the different conceptualisations of feature representations presented in table 4.2 (see §4.2.5.1), the Audio Features Ontology deals with data density, and temporal characteristics. It differentiates between dense, signal-like features of various dimensionality, for instance a chromagrams and detection functions, and sparse features that are scattered across the signal timeline, for instance, notes, or onsets.
92 6 Gyorgy Fazekas
93 6 Gyorgy Fazekas
** Does it serve us well?
94 1 Gyorgy Fazekas
** For example, loudness is defined as a segment in AF, and it does not fit a perceptual attribute well.
95 1 Gyorgy Fazekas
** What depth do we want ? (both in terms of scope and the level of detail in describing a feature extraction workflow)
96 1 Gyorgy Fazekas
** How AF relates to the DSP workflows used in extracting them?
97 1 Gyorgy Fazekas
98 1 Gyorgy Fazekas
99 1 Gyorgy Fazekas
h2. Existing resources :
100 1 Gyorgy Fazekas
101 2 Gyorgy Fazekas
h3. Some work related to Steve's use cases, segmentation and Ontologies:
102 1 Gyorgy Fazekas
103 2 Gyorgy Fazekas
** SALAMI Project: Kevin Page, DaveDeRoure http://salami.music.mcgill.ca/
104 2 Gyorgy Fazekas
** The Segment Ontology: http://users.ox.ac.uk/~oerc0033/preprints/admire2011.pdf
105 2 Gyorgy Fazekas
** PopStructure Ontology: Kurt Jacobson Unpublished. 
106 2 Gyorgy Fazekas
(Example available: http://wiki.musicontology.com/index.php/Structural_annotations_of_%22Can%27t_buy_me_love%22_by_the_Beatles) 
107 2 Gyorgy Fazekas
** Similarity Ontology: Kurt Jacobson http://grasstunes.net/ontology/musim/musim.html
108 1 Gyorgy Fazekas
109 1 Gyorgy Fazekas
110 2 Gyorgy Fazekas
h2. Ideas/resources for new Ontologies:
111 1 Gyorgy Fazekas
112 2 Gyorgy Fazekas
** Steve has worked on Acoustics related ontology
113 1 Gyorgy Fazekas
114 2 Gyorgy Fazekas
** Creating a DSP ontology:
115 2 Gyorgy Fazekas
** include processing steps down to math operations 
116 2 Gyorgy Fazekas
  (this can take advantage to the math:namespace in CWM: http://www.w3.org/DesignIssues/Notation3.html)
117 2 Gyorgy Fazekas
** describe common DSP parameters
118 2 Gyorgy Fazekas
119 2 Gyorgy Fazekas
** create an Acoustics Ontology
120 2 Gyorgy Fazekas
** describe Musicological concepts
121 2 Gyorgy Fazekas
** describe concepts related to cognitive and perceptual issues
122 2 Gyorgy Fazekas
123 2 Gyorgy Fazekas
124 1 Gyorgy Fazekas
h2. Currently missing features
125 1 Gyorgy Fazekas
126 2 Gyorgy Fazekas
** MFCC-s
127 1 Gyorgy Fazekas
** Rythmogram
128 1 Gyorgy Fazekas
** RMS energy
129 2 Gyorgy Fazekas
** combined features, e.g. weighted combinations or statistical averages over features
130 1 Gyorgy Fazekas
131 1 Gyorgy Fazekas
132 1 Gyorgy Fazekas
h2. Development issues
133 1 Gyorgy Fazekas
134 1 Gyorgy Fazekas
** chaining, combination, weighting
135 1 Gyorgy Fazekas
** how you associate features with arbitrary data
136 2 Gyorgy Fazekas
** summary feature types 
137 1 Gyorgy Fazekas
** SM (similarity matrix) are they part of the ontoogy?
138 2 Gyorgy Fazekas
** how to describe salience, can you hear it, can you perceive, is there an agreement
139 2 Gyorgy Fazekas
** how to describe weighting, confidence
140 1 Gyorgy Fazekas
** mood, music psychology, cognition, emotion, (perception ?)
141 1 Gyorgy Fazekas
** provenance => music provenance
142 2 Gyorgy Fazekas
** deprecation and versioning
143 1 Gyorgy Fazekas
144 1 Gyorgy Fazekas
145 1 Gyorgy Fazekas
h2. Long term objectives:
146 1 Gyorgy Fazekas
147 1 Gyorgy Fazekas
Some concrete tasks that can be done as the outcome of the collaboration:
148 1 Gyorgy Fazekas
149 2 Gyorgy Fazekas
** A version of Sonic Annotator that produces output adhering the new ontology
150 1 Gyorgy Fazekas
** Are we making people happier by doing so?
151 2 Gyorgy Fazekas
** gradual transition period?
152 2 Gyorgy Fazekas
** extend other software toolkits; e.g. a verison of Marsyas in C++
153 2 Gyorgy Fazekas
** multitrack processing using Sonic Annotator (this feature might come along soon)
154 1 Gyorgy Fazekas
155 1 Gyorgy Fazekas
156 2 Gyorgy Fazekas
h2. Some immediate tasks (TODO):
157 1 Gyorgy Fazekas
158 2 Gyorgy Fazekas
** collect more resources 
159 2 Gyorgy Fazekas
** Verify the relationship between AF as is, and other feature/segmentation Ontologies
160 2 Gyorgy Fazekas
** what other software uses it?
161 2 Gyorgy Fazekas
** papers and literature review
162 2 Gyorgy Fazekas
** relation to projects e.g. SIEMAC
163 2 Gyorgy Fazekas
** collect features that we need
164 2 Gyorgy Fazekas
** define scope (extend the diagram of the set of ontologies: )
165 2 Gyorgy Fazekas
** collect specific application examples from existing processing chain / workflow
166 1 Gyorgy Fazekas
167 2 Gyorgy Fazekas
collect software/projects that use/produce audio features:
168 1 Gyorgy Fazekas
169 1 Gyorgy Fazekas
** plugins, LADSPA, VAMP, Marsyas, CLAM, libextract, COMirva, MIRtoolbox, Supercollider, other frameworks
170 3 Gyorgy Fazekas
171 3 Gyorgy Fazekas
172 3 Gyorgy Fazekas
!http://www.isophonics.net/sites/isophonics.net/files/combined-frameworks.png!