-
Notifications
You must be signed in to change notification settings - Fork 2
Expand file tree
/
Copy pathresearch.html
More file actions
177 lines (168 loc) · 7.84 KB
/
Copy pathresearch.html
File metadata and controls
177 lines (168 loc) · 7.84 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
---
layout: default
title: Research
permalink: /research/
description: >-
The six tracks EthioNLP organises its work around, datasets, models,
applications, evaluation, capacity building and linguistics.
---
{%- assign hf = site.data.generated.huggingface -%}
{%- assign gh = site.data.generated.github -%}
{%- assign pubs = site.data.generated.publications.publications -%}
<div class="page-head">
<div class="wrap wrap--wide">
<p class="eyebrow">Research</p>
<h1 class="page-head__title">Research tracks</h1>
<p class="lede">
The tracks follow a sequence: text, then annotation, then a model, then
applications built on it, then evaluation, alongside the training of
people for each stage.
</p>
</div>
</div>
{%- for area in site.data.focus_areas -%}
{%- comment -%}
Same filter as /ecosystem/: focused breadth, ethionlp or member tier. Without
it the "open resources" slots below filled with bigscience/mt0-small and
masakhane/afrimgsm — the exact mistake /ecosystem/ documents at the top of the
file — and the "All N resources" link counted every artifact plus repos
against a destination page that renders neither.
{%- endcomment -%}
{%- assign ours = hf.artifacts
| where: "breadth", "focused"
| where_exp: "a", "a.tier == 'ethionlp' or a.tier == 'member'" -%}
{%- assign arts = ours | where: "focus", area.key -%}
{%- assign repos = gh.repos | where: "focus", area.key -%}
{%- comment -%}
Members carry `focus: []` by default; the real value is derived by
derive_focus.py and reached through focus-of.html, exactly as /community/ does
it. Filtering site.members on m.focus directly matched nobody, so this panel
rendered on none of the six tracks.
{%- endcomment -%}
{%- comment -%}
Canonical community order, the `order` field on each member file -- the same
sequence /community/, /mentorship/ and people-ordered.html use. Iterating
site.members unsorted gave Jekyll's own order, which for these files is their
slug, so the panel read as an alphabetical list. Members imported rather than
founded carry a high order number and so land at the end, which is where
later joiners belong.
{%- endcomment -%}
{%- assign roster = site.members | sort: "order" -%}
{%- assign people = "" | split: "" -%}
{%- for m in roster -%}
{%- include focus-of.html m=m -%}
{%- if focus contains area.key -%}{%- assign people = people | push: m -%}{%- endif -%}
{%- endfor -%}
{%- assign alt = forloop.index0 | modulo: 2 -%}
<section class="band{% if alt == 1 %} band--alt{% endif %}" id="{{ area.key }}">
<div class="wrap wrap--wide">
<div class="split split--sidebar">
<div>
<span class="focus__cell-icon" style="--focus-color: var(--{{ area.color }}); --focus-tint: var(--{{ area.color }}-soft); margin-bottom: var(--sp-4)">
{% include icon.html name=area.icon %}
</span>
<h2>{{ area.title }}</h2>
<p class="lede">{{ area.summary }}</p>
{%- if area.examples %}
<ul style="margin-top: var(--sp-4)">
{%- for ex in area.examples %}<li>{{ ex }}</li>{% endfor %}
</ul>
{%- endif %}
{%- if arts.size > 0 or repos.size > 0 %}
<h3 style="margin-top: var(--sp-6); font-size: var(--t-md)">Open resources in this track</h3>
<div class="grid grid--2 grid--tight">
{%- for a in arts limit: 4 %}{% include artifact-card.html a=a %}{% endfor %}
{%- for r in repos limit: 2 %}{% include repo-card.html r=r %}{% endfor %}
</div>
{%- comment -%}
Both halves of a track: what has been released, and what has been
written. The publications link carries the same ?track= the facet on
/publications/ reads, so the two stay in step.
{%- endcomment -%}
{%- assign track_pubs = site.data.generated.publications.publications
| where_exp: "p", "p.tracks contains area.key" -%}
{%- if track_pubs.size > 0 %}
<p style="margin-top: var(--sp-4)">
<a href="{{ '/publications/?track=' | append: area.key | relative_url }}">
{{ track_pubs | size }} paper{% unless track_pubs.size == 1 %}s{% endunless %} in {{ area.title | downcase }} →
</a>
</p>
{%- endif %}
{%- comment -%}
Repos are deliberately not added in: /ecosystem/ renders artifacts only,
so counting repositories here sent the reader to a page showing fewer
items than the link promised. Suppressed entirely at zero, which is what
evaluation, capacity and linguistics are — they each linked to
"Nothing matches those filters."
{%- endcomment -%}
{%- if arts.size > 0 %}
<p style="margin-top: var(--sp-4)">
<a href="{{ '/ecosystem/?focus=' | append: area.key | relative_url }}">
All {{ arts.size }} resource{% unless arts.size == 1 %}s{% endunless %} in {{ area.title | downcase }} →
</a>
</p>
{%- endif %}
{%- endif %}
</div>
<aside class="stack">
{%- if people.size > 0 %}
<div class="card">
<h3 class="card__title" style="font-size: var(--t-md)">Who works on this</h3>
<ul style="list-style: none; padding: 0; margin: 0; font-size: var(--t-sm)">
{%- for m in people limit: 8 %}
<li style="margin-bottom: .45em"><a href="{{ m.url | relative_url }}">{{ m.name }}</a></li>
{%- endfor %}
</ul>
<div class="card__foot">
{%- comment -%}
The list is capped at 8, so say how many there are: otherwise a
track with 15 people looks like a track with 8, and the members at
the end of the canonical order are invisible from here.
{%- endcomment -%}
<a class="btn btn--sm btn--ghost btn--block" href="{{ '/community/?focus=' | append: area.key | relative_url }}">
{% if people.size > 8 %}See all {{ people.size }}{% else %}See all{% endif %}
</a>
</div>
</div>
{%- endif %}
<div class="card">
<h3 class="card__title" style="font-size: var(--t-md)">Joining this track</h3>
<div class="card__foot">
<a class="btn btn--sm btn--block" href="{{ '/join/' | relative_url }}">Get involved</a>
</div>
</div>
</aside>
</div>
</div>
</section>
{%- endfor -%}
<!-- ── Mission ───────────────────────────────────────────────────────────── -->
<section class="band band--ink">
<div class="wrap wrap--wide">
<div class="split">
<div>
<p class="eyebrow" style="color: rgba(252,251,249,.6)">Why EthioNLP</p>
<h2>Origins and goals</h2>
<p class="lede" style="color: rgba(252,251,249,.8)">
There are efforts throughout the world to conduct NLP for Ethiopian
languages, but without formal communication among researchers. At
COLING 2018 in Santa Fe, researchers from Addis Ababa
University, the University of Hamburg and the University of Trento
took the initiative to create a formal society for Ethiopian-language
NLP research.
</p>
</div>
<div>
<h3 style="font-size: var(--t-md)">Our goals</h3>
<ul style="color: rgba(252,251,249,.82)">
<li>Identify, prioritise and focus NLP research topics for Ethiopian languages.</li>
<li>Organise workshops, seminars and conferences for Ethiopic NLP research.</li>
<li>Join efforts among Ethiopian NLP researchers around the world.</li>
<li>Support and mentor students in NLP and data science research.</li>
<li>Coordinate resources for Ethiopian language research.</li>
<li>Collaborate with Ethiopian universities and assist education quality.</li>
</ul>
</div>
</div>
</div>
</section>